# Agents
Source: https://docs.atomicagi.com/automation/agents/overview
Build, manage, and improve agents so every recurring marketing task gets the right specialist
![Agents overview with the chat launcher, All agents action, and chat history]()
Use this page to control which agent should do each type of work. This is where you standardize output quality across your team.
## Questions this page should answer
1. Which specialist should run this task?
2. Do we already have an agent for this use case?
3. Which agents need updates because outputs are weak or inconsistent?
## Before you manage agents
* Define your top use cases first: analysis, technical SEO, content, automation.
* Keep one clear purpose per agent.
* Use simple names based on outcome, not internal jargon.
## What this page gives you
* `All agents` page for full agent discovery.
* `My agents` workspace to manage custom agents.
* AI sidebar access using the top-right 3-boxes AI assistant button.
* Chat history to review real usage before editing instructions.
## All agents page
Open `All agents` from the top-right actions on the Agents overview when you want to pick the right specialist quickly.
![All agents page with available specialists and pagination]()
How to use it:
* Review the available agents and compare their descriptions.
* Use the creation date and chat count to understand how each agent is used.
* Select an agent to start a new chat with that specialist.
* Return to the Agents overview when you want to reuse a recent conversation.
## AI sidebar (quick actions on Agents)
Use this when you want to stay on the Agents page while running quick AI tasks.
1. From the Agents page, use the top-right 3-boxes AI assistant button.
2. The AI sidebar opens on the right for prompts and quick actions.
3. Keep using the same Agents view while running analysis or content tasks.
![AI sidebar open on the right side of the Agents page for quick prompts and actions]()
## My agents page (management workspace)
This is where you maintain your custom agents.
![My agents page with custom agent cards and create action]()
Use `My agents` to:
* Review your custom agent list.
* Spot duplicates and overlap.
* Identify low-value agents to merge or retire.
* Decide which agents need instruction updates.
Practical rule: if two agents solve the same problem, keep one and improve it.
## Create agent flow (new specialist)
Create a new agent only when an existing one cannot be improved to cover the task.
![Create agent form with name, description, instructions, and use cases]()
What to fill carefully:
* `Name`: outcome-focused and easy to route.
* `Description`: one sentence on what this agent produces.
* `Instructions`: behavior, boundaries, and response format.
* `Use cases`: concrete quick actions users can click.
* `Allowed tool categories`: limit tools to reduce noisy or risky output.
A good instruction block should include:
1. What the agent should do
2. What it should avoid
3. Expected output format
4. Decision rules when data is incomplete
Important: Keep one clear job per agent. Multi-purpose
agents create inconsistent output and are harder to maintain.
## Edit agent flow (quality improvement)
Editing is where most quality gains happen.
![Agent editing context from My agents for refining behavior and scope]()
When to edit an agent:
* Outputs are too generic.
* Answers are correct but not actionable.
* Team members use long prompt workarounds to get useful results.
* The same mistakes appear across multiple runs.
How to improve fast:
1. Review 5-10 recent runs for repeated failure patterns.
2. Tighten instructions with explicit output structure.
3. Add or remove tool access based on mistakes.
4. Re-test using the same prompt set and compare results.
## Weekly agent operations routine
1. Review top-used agents and weakest outputs.
2. Update one high-impact agent each week.
3. Remove redundant agents.
4. Promote proven prompts into reusable workflows.
## Keep in mind
* More agents do not mean better operations.
* Clear scope beats “do everything” instructions.
* Most performance issues come from vague instructions, not model quality.
## Where to go next
* [New chat](/automation/new-chat/overview): run requests with the right agent selected
* [Teams](/automation/teams/overview): group specialists into reusable multi-agent setups
* [Tasks](/automation/tasks/overview): track work assigned to agents and people
* [Workflows](/automation/workflows/overview): automate repeated agent tasks
* [Automations](/automation/automations/overview): schedule recurring agent conversations
# Automations
Source: https://docs.atomicagi.com/automation/automations/overview
Create recurring agent conversations from ready-made templates so scheduled analysis and follow-up stay consistent
![Automations page with recurring agent conversation templates and automation cards]()
Use Automations to schedule recurring agent conversations. This page is for repeatable report, prep, and triage prompts that should run on a cadence without manual kickoff.
## Questions this page should answer
1. Which recurring automations already exist?
2. Which template should we use for the next recurring job?
3. Which automation should be paused, edited, run now, or removed?
## Before you automate
* Run the prompt manually in [New chat](/automation/new-chat/overview) before scheduling it.
* Confirm the output has a clear owner and review rhythm.
* Decide whether this should be an Automation or a full [Workflow](/automation/workflows/overview).
* Keep the first version narrow: one recurring question, one expected output, one cadence.
## What this page gives you
* Template groups for recurring report, content, and triage jobs.
* Cadence defaults for each template.
* Automation cards with run, edit, pause, and delete actions.
* Detail pages for the prompt, schedule, model, status, and previous runs.
## How to read the automation library
Read each automation card as an operating commitment:
* `Name`: what recurring question or output this automation owns.
* `Template`: the starting pattern used to create the automation.
* `Cadence`: how often the conversation will run.
* `Status`: whether the automation is active, paused, or needs attention.
* `Previous runs`: whether the scheduled output is staying useful over time.
Use this rule:
* If the output needs multiple tools, branching, or structured handoffs, use a Workflow.
* If the output is a recurring agent conversation with a stable prompt, use an Automation.
* If the prompt still changes every time you run it, keep it in New chat until it stabilizes.
## How to use this page
### Start from the closest template
Templates reduce setup mistakes. Pick the nearest recurring use case first, then tighten the prompt instead of starting from a blank automation every time.
### Keep cadence proportional to decision value
Daily or every-48-hour runs only make sense when the output creates real decisions. Low-value prompts should stay weekly or monthly.
### Review previous runs before editing
If an automation underperforms, inspect the last outputs first. Weak recurring output usually needs a better prompt, tighter tool scope, or a lower cadence.
## Quick weekly checklist
1. Review failed, paused, or stale automations first.
2. Check whether active automations still match the current reporting rhythm.
3. Open recent runs and confirm the output is still actionable.
4. Pause automations that no one reads or uses.
5. Convert repeatable multi-step automations into workflows when they outgrow a single prompt.
## What to fix first
| Pattern in Automations | What it usually means | Recommended action |
| ------------------------------------ | --------------------------------------- | --------------------------------------------------- |
| Many active automations, few readers | Cadence is too noisy | Pause low-value runs and keep only decision outputs |
| Same prompt edited before every run | Automation was created too early | Move back to New chat until the prompt stabilizes |
| Output is useful but incomplete | Prompt lacks context or constraints | Tighten the prompt and rerun manually first |
| Automation needs several handoffs | It has become an operational workflow | Rebuild as a Workflow with explicit steps |
| Failures repeat across runs | Tool, data, or permission setup is weak | Fix the underlying setup before re-enabling |
## Team routine
1. Weekly: review recent runs and pause anything not used.
2. Bi-weekly: promote proven automations into team operating routines.
3. Monthly: delete obsolete automations and document the few that matter.
## Keep in mind
* Automations are scheduled agent conversations, not full workflow graphs.
* A bad prompt becomes expensive faster when it is scheduled.
* Run a manual version first if you are not sure the prompt is stable.
## Where to go next
* [New chat](/automation/new-chat/overview)
* [Agents](/automation/agents/overview)
* [Tasks](/automation/tasks/overview)
* [Workflows](/automation/workflows/overview)
# Grids
Source: https://docs.atomicagi.com/automation/grids/overview
Organize row-by-row work in spreadsheet-style project workspaces
![Grid workspace with Title, Keyword, Review date, and Priority columns]()
Use Grids when a task starts with a list: keywords, URLs, pages, products, or briefs. A Grid keeps those items in rows so you can add simple fields, sort columns, and review work in one project workspace.
Grids are available on Team and higher plans. On Free or Starter, opening Grids shows the Team upgrade option.
## Questions this page should answer
1. Which list should become rows?
2. Which fields should become columns?
3. Which rows need review, cleanup, or follow-up?
## Before you create a grid
* Start with one clear list of inputs.
* Decide what the `Title` column should represent.
* Add only the columns you need for the next decision.
* Use separate sheets when two lists need different structure.
## What this page gives you
* Project-level Grid workspaces.
* Sheets inside each Grid.
* Five starter rows and one starter `Title` column.
* Text, number, date, content, JSON, URL, image URL, and workflow columns.
* Inline cell editing.
* Sheet tools for creating, importing, reordering, renaming, duplicating, and deleting sheets.
* Row tools for selecting, deleting, and dragging rows into a new order.
* Column tools for inserting, renaming, changing type, sorting, and deleting columns.
* CSV export for the active sheet.
## Create a grid
Open `Automation > Grids`, then use `New Grid`.
Each new Grid starts with one sheet named `Sheet 1`, one `Title` column, and five empty rows.
Use the in-table `Add Column` header and `Add Row` row to expand the active sheet.
## Manage sheets
![Sheet menu with rename, duplicate, and delete controls]()
Use the sheet tabs at the top of the workspace to switch sheets. Drag a sheet tab to change the sheet order.
Click `+` in the tab strip to add another sheet. You can start with a blank sheet, import a public Google Sheet, import a CSV file, import URLs from a sitemap, or create a sheet from a workflow.
![Create sheet menu with Blank, From Workflow, From Google Sheets, From CSV file, and From Sitemap options]()
Google Sheets imports use the first row as column headers and create one row for each remaining spreadsheet row. Use a public Google Sheets share link so Atomic can read the sheet export.
Open a sheet menu when you need to rename, duplicate, or delete a sheet. Delete is only available when the Grid has more than one sheet.
## Add columns
![Add Column menu with Text, Number, Date, and Workflow run column types]()
Use `Add Column` for row data such as:
* Text
* Number
* Date
* Content
* JSON
* URL
* Image URL
Use text for names, keywords, notes, and free-form values. Use number for priority, score, volume, or counts. Use date when the row needs a due date, review date, or publish date.
Use `Content` for longer writing. It opens in a full-screen editor. Use `JSON` for structured values. Invalid JSON stays editable until you fix it. Use `URL` for links. Valid links show an open-in-new-tab action. Invalid links show a warning. Use `Image URL` when the cell should show an image preview from a URL.
Use `Workflow run` when each row should run the same workflow with row-specific inputs.
![Workflow run column configuration drawer with workflow selection and input mapping]()
When you choose `Workflow run`, Grids opens a side drawer. Name the column, choose a workflow, then map the workflow inputs to existing columns, new columns, hardcoded values, or empty optional inputs.
To change a workflow column later, open its column header menu and select `Edit inputs and outputs`. The drawer keeps the selected workflow fixed, shows its current outputs, and lets you remap each input without rebuilding the column.
![Edit workflow column drawer with a hardcoded input and the workflow output list]()
## Enter row values
Click into a cell and type the value. The cell saves when you leave it.
Use `Add Row` at the bottom of the table when the starter rows are full. Keep one idea, page, keyword, or task per row so sorting stays useful.
Hover over a row number to show row controls. Use the checkbox to select rows. Selected rows show a delete action at the bottom of the workspace. Use the drag handle to reorder rows.
## Manage columns
![Grid column menu with rename, type, sort, and delete controls]()
Click a column header to rename it, insert a column to the left or right, change the column type, sort rows by that column, or delete the column.
Changing a column type changes how the cells in that column are edited. Check existing values after changing type, especially when moving between text, number, and date.
## Export a sheet
Use `Export CSV` to download the active sheet. The export uses the current sheet columns and row order. Values with commas, quotes, or line breaks are escaped for CSV.
## Quick routine
1. Create a Grid for one list.
2. Add the columns needed to review that list.
3. Fill the first few rows.
4. Sort by priority, date, or another decision column.
5. Select and delete rows that no longer belong.
6. Export CSV when you need to review or share the active sheet outside Atomic.
7. Add a new sheet when the next list needs different columns.
## Keep in mind
* Grids are project-level workspaces.
* New Grids and blank sheets start with one `Title` column and five rows.
* Google Sheets, CSV, and sitemap imports create a new sheet from the linked, uploaded, or discovered data.
* Keep column names short so the table stays easy to scan.
* Delete removes the selected rows, sheet, or column and its data.
## Where to go next
* [New chat](/automation/new-chat/overview): ask an agent to help plan or populate a Grid
* [Agents](/automation/agents/overview): choose agents for repeatable project work
* [Tasks](/automation/tasks/overview): turn grid output into managed follow-up
* [Workflows](/automation/workflows/overview): build reusable workflows for structured work
# New chat
Source: https://docs.atomicagi.com/automation/new-chat/overview
Start focused analysis quickly, pick the right agent, and turn prompts into actions
![New chat composer with agent and chat selectors]()
Use New chat when you need a fast answer and a clear next step. This page helps you go from question to action in minutes.
## Questions this page should answer
1. Which agent should handle this request?
2. What prompt format gets an actionable answer?
3. How do you keep context clean across different tasks?
## Before you start
* Pick one clear goal for this chat.
* Know what output you need: diagnosis, action plan, or draft copy.
* Choose whether this should stay ad-hoc or become a workflow later.
## What this page gives you
* A central prompt composer for one-off requests.
* An agent selector to match task type with the right specialist.
* A chat-context selector to continue an old thread or start clean.
* Quick prompt starters for common marketing tasks.
## How the composer is organized
* Main input box: write the task in plain language.
* Agent selector: pick the specialist that should answer.
* Chat selector: choose `New chat` when you want clean context.
* Quick starter chips: launch common requests quickly.
Start with a new chat for analysis work, then continue the thread only if the follow-up is part of the same task.
## Choose the right agent before sending
![New chat with General agent dropdown open]()
Agent choice changes output quality more than prompt wording.
* `General agent`: broad planning and mixed questions.
* `SEO performance analyst`: traffic and ranking diagnosis.
* `Technical SEO auditor`: crawl, index, and issue-focused tasks.
* `Content strategist` or `Content editor`: content planning and rewriting.
* `Workflow automation specialist`: requests that should be turned into repeatable flows.
Important: If output quality is weak, switch the agent
first before rewriting the prompt.
If the result feels too generic, switch agent first, then retry the same prompt.
## Prompt format that works for marketers
Use this structure:
1. Goal
2. Context
3. Constraints
4. Output format
Example:
```text theme={null}
Goal: Find my top 5 pages losing organic traffic in the last 28 days.
Context: B2B SaaS, focus on non-branded traffic.
Constraints: Prioritize quick wins we can ship this sprint.
Output: Table with page, issue diagnosis, and next action.
```
## When to keep chatting vs start fresh
Start a fresh chat when:
* You switch from SEO analysis to content writing.
* You change site, market, or campaign context.
* You want a clean answer without prior assumptions.
Stay in the same chat when:
* You are refining the same output.
* You are asking follow-up implementation questions.
* You are validating tradeoffs for the same problem.
## Quick execution checklist
1. Pick agent.
2. Write prompt with clear output format.
3. Review answer for direct actions.
4. Convert useful repeated tasks into workflows.
## Keep in mind
* Better inputs produce better outputs.
* One task per prompt keeps answers focused.
* New chat is best for speed; workflows are best for consistency.
## Where to go next
* [Agents](/automation/agents/overview): create and manage specialists
* [Teams](/automation/teams/overview): reuse multi-agent setups for bigger jobs
* [Workflows](/automation/workflows/overview): automate repeatable tasks
* [Automations](/automation/automations/overview): schedule recurring agent conversations
# Schedule
Source: https://docs.atomicagi.com/automation/schedule/overview
Set recurring SEO audit and workflow schedules with clear timing and notifications
![Schedule page with SEO audit scheduling controls]()
Use Schedule to keep recurring execution on time without manual follow-up. This page controls when audits and workflows run.
## Questions this page should answer
1. Are critical checks running at the right cadence?
2. Are schedules aligned to team timezone and reporting rhythm?
3. Are stakeholders receiving completion alerts reliably?
## Before you configure schedules
* Confirm your reporting cadence (daily, weekly, bi-weekly, monthly).
* Align timezone with the team that receives outputs.
* Decide which runs need email notifications.
## What this page gives you
* Separate scheduling for SEO audits and workflows.
* Cadence controls from daily to inactive.
* Time window and timezone controls.
* Notification toggle for completion emails.
## SEO Audit tab
This tab controls recurring technical/site audit runs.
![SEO Audit tab in Schedule with cadence and notification settings]()
Use this tab when you want predictable technical monitoring.
* Daily: active monitoring for high-change sites.
* Weekly: standard cadence for most teams.
* Bi-weekly or monthly: lighter oversight for stable sites.
* Inactive: pause runs during migrations or maintenance windows.
## Workflows tab
This tab controls recurring workflow automations.
![Workflows tab in Schedule for automation timing]()
Use this tab for repeatable operational outputs:
* Weekly priority lists for team planning.
* Bi-weekly update workflows for content refresh cycles.
* Monthly executive summaries and performance snapshots.
## Cadence selection guide
* Choose `Daily` when timing is business-critical.
* Choose `Weekly` for most SEO/content operating rhythms.
* Choose `Bi-weekly` when work cycles are sprint-based.
* Choose `Monthly` for leadership rollups and trend reporting.
If output quality drops, reduce frequency and improve workflow logic before scaling run volume.
Important: Set schedule only after manual runs are stable.
Scheduling an unstable workflow only repeats bad output faster.
## Notification hygiene
* Enable completion emails only for workflows that require action.
* Avoid alert fatigue by limiting low-priority notifications.
* Route critical failures to a shared team inbox.
## Weekly schedule review checklist
1. Verify next-run timing is correct.
2. Check timezone after team/location changes.
3. Confirm completion notifications are still useful.
4. Pause obsolete schedules and remove stale automations.
## What to fix first
| Pattern in Schedule | What it usually means | Recommended action |
| ------------------------------- | ------------------------------------ | -------------------------------------------- |
| Scheduled output is ignored | Cadence or audience is wrong | Reduce frequency or route to a clearer owner |
| Runs happen at the wrong time | Timezone or day setting is stale | Update time settings and verify the next run |
| Workflow output is inconsistent | Workflow is not ready for scheduling | Return to manual runs and fix workflow logic |
| SEO audit fires too often | Monitoring exceeds decision cadence | Move to weekly or monthly |
| Notifications create noise | Alerts are not tied to action | Disable low-value completion emails |
## Team routine
1. Weekly: review which scheduled outputs produced decisions.
2. Bi-weekly: tune cadence for workflows tied to active campaigns.
3. Monthly: pause schedules that no longer match team operations.
## Keep in mind
* Correct cadence is part of quality, not just convenience.
* Over-scheduling low-value runs wastes credits and attention.
* A schedule should match decision cadence, not tool availability.
## Where to go next
* [Workflows](/automation/workflows/overview): improve what gets scheduled
* [Automations](/automation/automations/overview): manage recurring agent conversations separately
* [Technical overview](/data/technical/overview): interpret scheduled audit outputs
* [Agents](/automation/agents/overview): route scheduled tasks to the right specialists
# Tasks
Source: https://docs.atomicagi.com/automation/tasks/overview
Track work assigned to you, specific agents, or the general agent so follow-through stays visible
![Tasks page with status filters, assignee filters, and task list area]()
Use Tasks to turn AI output into managed follow-through. This page is the work queue for manual tasks, general-agent tasks, and tasks assigned to named agents.
## Questions this page should answer
1. What is open, blocked, or done right now?
2. Who owns each task?
3. Which agent or conversation produced the work?
## Before you triage
* Start with the team’s current operating window: today, this week, or sprint.
* Decide whether you are reviewing human-owned work, agent-owned work, or both.
* Confirm task status definitions are used consistently by the team.
* Open the originating conversation when task context is unclear.
## What this page gives you
* List and board views.
* Status filters for `Open`, `In progress`, `Blocked`, `Done`, and `Canceled`.
* Assignee filters for you, the general agent, unassigned work, or specific agents.
* Task detail, checklist progress, and conversation links.
## How to read the task queue
Review the page in this order:
1. `Blocked`: work that needs a decision, missing context, or reassignment.
2. `In progress`: active work that may need owner confirmation.
3. `Open`: work that has not started.
4. `Done` and `Canceled`: completed history and cleanup signals.
Use this rule:
* A task with no owner should be routed before it is refined.
* A blocked task should include the specific blocker, not only a blocked status.
* A task created from a weak agent output should trigger prompt or agent review.
## How to use this page
### Triage by status first
Review blocked and in-progress tasks before opening new work. A stale blocked queue usually means outputs are not turning into decisions.
### Check assignment mode
Some tasks belong to you. Others belong to a named agent or to the general agent. Use that distinction to decide whether the next step is execution, reassignment, or prompt cleanup.
### Use board view for flow, list view for detail
Board view is better for active movement across statuses. List view is better when you need identifiers, due dates, or denser review.
## Quick weekly checklist
1. Clear stale blocked tasks first.
2. Reassign unowned work to a person, named agent, or the general agent.
3. Close done work that no longer needs review.
4. Open conversation-linked tasks with unclear context.
5. Flag repeated weak tasks back to the agent or workflow that created them.
## What to fix first
| Pattern in Tasks | What it usually means | Recommended action |
| ---------------------------- | ----------------------------------------- | ------------------------------------------------- |
| Blocked column keeps growing | Owners lack context or decision authority | Add the blocker, assign a decision owner |
| Many unassigned tasks | Routing rules are unclear | Assign owners and review task creation prompts |
| Done tasks remain unreviewed | Completion criteria are vague | Add explicit finish criteria to future tasks |
| Repeated duplicate tasks | Upstream automation is too broad | Tighten the automation, workflow, or agent prompt |
| Agent tasks stall frequently | Agent role or tool access is wrong | Review the agent setup and available tools |
## Team routine
1. Daily: scan blocked and in-progress work.
2. Weekly: clean stale open tasks and review repeated failure patterns.
3. Monthly: audit whether task categories match how the team actually works.
## Keep in mind
* Too many unassigned tasks usually point to unclear routing.
* A growing blocked column usually means the upstream prompt or team needs work.
* Task quality depends on clear owners and clear finish criteria.
## Where to go next
* [Agents](/automation/agents/overview)
* [Teams](/automation/teams/overview)
* [Automations](/automation/automations/overview)
* [Workflows](/automation/workflows/overview)
# Teams
Source: https://docs.atomicagi.com/automation/teams/overview
Group agents into reusable teams so multi-step work has a stable orchestrator and clear specialists
![Teams page showing reusable AI teams in the automation section]()
Use Teams when one agent is not enough. This page helps you define reusable multi-agent groups for recurring work that needs planning, execution, and review roles.
## Questions this page should answer
1. Which agents should work together as a team?
2. Which team already fits this workflow?
3. Which teams need role or member changes?
## Before you create a team
* Define the repeated workflow before choosing agents.
* Confirm one agent should coordinate the work.
* Decide what each specialist contributes and where handoff happens.
* Avoid creating a team when a single agent can produce the output reliably.
## What this page gives you
* A dedicated team library alongside the agent library.
* Team detail and edit flows.
* A stable way to reuse the same team structure across future tasks.
## How to read the team library
Read each team as a reusable operating pattern:
* `Team name`: the outcome or workflow the team supports.
* `Members`: the specialist agents involved.
* `Coordinator`: the agent responsible for planning and handoff.
* `Description`: the boundary of work the team should own.
Use this rule:
* Use a team when work needs planning, execution, and review roles.
* Use one agent when the task is narrow and does not need role separation.
* Use a workflow when the steps, inputs, and outputs need stronger structure.
## How to use this page
### Create teams around a workflow, not around departments
Good teams map to a real operating pattern such as research plus writing plus QA. They should exist because the same sequence repeats often enough to deserve reuse.
### Keep one orchestrator
The team should have one clear coordinator. Without that, the output quality becomes hard to predict and task handoffs become noisy.
### Review teams after repeated failures
If a workflow stalls or produces inconsistent results, review the team roles before adding more prompt instructions.
## Quick monthly checklist
1. Review teams that were used in the last month.
2. Remove teams that duplicate a simpler single-agent setup.
3. Check whether each team has one clear coordinator.
4. Update roles when workflows or strategy change.
5. Test important teams with a fresh chat before relying on them in a workflow.
## What to fix first
| Pattern in Teams | What it usually means | Recommended action |
| ---------------------------------- | ------------------------------------ | ------------------------------------------------ |
| Several agents do the same role | Team structure is redundant | Remove overlap or merge roles |
| No clear coordinator | Handoffs will be inconsistent | Assign one orchestrating agent |
| Team output is generic | Roles are too broad | Narrow each member’s purpose and expected output |
| Team only gets used once | It may not deserve a reusable team | Use New chat or a single agent instead |
| Repeated failures across workflows | Team roles do not match the workflow | Update team composition before changing prompts |
## Team routine
1. Before launch: test the team on one representative request.
2. After failures: review roles before adding more instructions.
3. Quarterly: prune unused or overlapping teams.
## Keep in mind
* Teams are for repeatability, not novelty.
* A strong team usually needs fewer prompt workarounds.
* If one agent can do the job well, keep the simpler setup.
## Where to go next
* [Agents](/automation/agents/overview)
* [Tasks](/automation/tasks/overview)
* [Workflows](/automation/workflows/overview)
* [New chat](/automation/new-chat/overview)
# Tools
Source: https://docs.atomicagi.com/automation/tools/overview
Inspect tool inputs, run tools directly, and understand which tool categories are available for the current project
![Tools page with categorized tool cards and direct execution support]()
Use Tools when you need to inspect or run an available tool directly. This page is the lowest-level view of the project tool surface.
## Questions this page should answer
1. Which tools are available in this project?
2. What inputs does a tool require?
3. What result shape should I expect before I wire the tool into an agent or workflow?
## Before you run tools directly
* Know the project and data source the tool should use.
* Confirm you have permission to run the tool.
* Start with the smallest valid input.
* Use direct runs for verification, not as a replacement for repeatable workflows.
## What this page gives you
* A categorized tool library.
* Tool descriptions and input schemas.
* Direct execution with form-based arguments.
* Raw results you can inspect or copy.
## How to read tool details
Read the tool panel before running anything:
* `Description`: what the tool is meant to do.
* `Input schema`: required fields, optional fields, and accepted value types.
* `Category`: where the tool fits in the project capability surface.
* `Result`: the raw shape an agent or workflow will receive.
Use this rule:
* If the tool succeeds here but an agent fails, the issue is likely prompt/context.
* If the tool fails here, fix credentials, permissions, or input shape first.
* If the output shape is too raw for a user, wrap it in an agent or workflow.
## How to use this page
### Start with the category
Use the category grouping to narrow the surface before you inspect individual tools. This is faster than searching every tool name when you only need one domain such as SEO, reports, or AI search.
### Read the required fields carefully
The input schema tells you which arguments are mandatory and what type each argument expects. Validate that first before assuming a tool is broken.
### Use direct runs for verification
Run the tool here when you need to confirm output shape, test a narrow input, or debug why an agent is producing weak results.
## Quick debugging checklist
1. Confirm the tool exists in the expected category.
2. Check required input fields before running.
3. Run with one narrow input.
4. Inspect the raw result shape.
5. Compare direct output with what the agent or workflow produced.
## What to fix first
| Pattern in Tools | What it usually means | Recommended action |
| -------------------------------- | ---------------------------------------- | --------------------------------------------- |
| Tool is missing | Feature, permission, or setup is absent | Check project settings and member permissions |
| Required input is unclear | Workflow or prompt needs a clearer value | Define the exact input before running |
| Direct run fails | Tool setup or data access is broken | Fix credentials, integration, or data source |
| Direct run succeeds, agent fails | Agent context is weak | Improve the agent prompt or tool instructions |
| Output is correct but noisy | Result needs interpretation | Use an agent/workflow step to summarize it |
## Team routine
1. Use Tools to verify new capabilities before adding them to agents.
2. Save known-good inputs for recurring workflows.
3. Re-test tools after integration or permission changes.
## Keep in mind
* This page is for verification and targeted runs, not for broad workflow orchestration.
* Tool availability depends on project capabilities and permissions.
* If a tool run is noisy, fix the input shape before changing the agent.
## Where to go next
* [Agents](/automation/agents/overview)
* [Workflows](/automation/workflows/overview)
* [Tasks](/automation/tasks/overview)
* [Automations](/automation/automations/overview)
# Workflows
Source: https://docs.atomicagi.com/automation/workflows/overview
Build, run, monitor, and schedule repeatable automations for your marketing operations
![Workflows library with workflow cards and run history table]()
Use Workflows to turn repeated tasks into reliable operations. This page helps you create the workflow, run it, debug it, and schedule it.
## Questions this page should answer
1. Which recurring task should become a workflow first?
2. Are runs completing with useful outputs, not just green status?
3. Where should we fix steps, inputs, or schedule to improve reliability?
## Before you automate
* Start with one repeated task your team already does manually.
* Define the exact output you expect before building.
* Keep version one short and clear.
## What this page gives you
* A workflow library with reusable cards.
* Fast creation from templates or from scratch.
* A full editor for step logic and sequencing.
* Per-workflow run, execution, and history views.
* Schedule controls for recurring runs.
## How to read the workflow library
In the top section:
* Each card is one automation.
* The card title and subtitle tell you the business job.
* `Last` or `Never run` tells you adoption and recency.
* The three-dot action on cards helps with quick maintenance actions.
Use the table below cards as your global operations feed:
* `Status`: run health.
* `AI credits`: cost per run.
* `Inputs`: confirms run context.
* `Started` and `Duration`: speed and timing consistency.
When a run fails, inspect `Inputs` first, then step output.
## Create workflow (template or scratch)
![Create workflow modal with start from scratch and template options]()
Use `Create workflow` to start from:
* `Start from scratch` for custom logic.
* Templates for common execution patterns.
Template-first is faster for most teams. Start from scratch only when your flow is unique.
## Edit workflow (builder)
![Workflow editor with visual step graph and step palette]()
The editor is where you design and refine execution:
* Left panel: available step types.
* Center canvas: step order and flow.
* Right panel: AI assistant and properties.
* Top actions: run test, save, and chat-guided edits.
Keep steps explicit. Each step should do one clear job.
### Content image enrichment workflows
Use content enrichment steps when a draft already exists and needs useful inline visuals:
* `Enrich content images`: search the project's Product Images library first, then use AI web search when no first-party image strongly matches. It adds the strongest images near the sections they support and saves the result as a new active editor version. It skips webpage screenshots.
* `Create content version`: save markdown from another step as a new active editor version without overwriting older versions.
Product Images keep their existing Atomic storage URL and do not need a public source-credit line. External images are imported into Atomic storage and include a visible credit to the original source.
This works well after `Generate article` or when refreshing older text-only drafts. Add clear titles, descriptions, tags, and use cases on [Product Images](/settings/project/product-images) so the agent can find the right first-party visual. Keep `target image count` balanced; weak visual opportunities should be skipped instead of forcing filler images.
### Repurpose content for distribution
Use `Repurpose content` after `Get content from editor` to adapt one finished article into one selected format:
* LinkedIn post or article
* X post or thread
* Newsletter issue
The step applies the current project Brand kit and can use explicitly selected Knowledge Bases for supporting project facts. No Knowledge Base selection means no project Knowledge search. It preserves source claims, numbers, quotes, attribution, and links instead of researching and rewriting the topic from scratch.
`Repurpose content` only returns a structured draft. It does not publish, schedule, authenticate, or create a draft on an external platform. Place `Human review` after it to approve or edit the draft. The `Repurpose content for distribution` template already uses this safe generation-and-review sequence and ends with the approved draft as workflow output.
For X, choose `Post` for one complete idea or `Thread` for ordered segments. If a faithful point cannot fit within the configured X post limit, the step skips the post instead of shortening it into a misleading teaser.
For a newsletter issue, choose a publication profile and provide the issue number. Previous-issue details, referral campaigns, commercial links, and other changing values are included only when you supply them. The `The Startup Finance` profile keeps its navigation, issue shell, rundown, TL;DR, numbered sections, mid-issue CTA position, bottom line, reply CTA, referral module, signature, and P.S. in one reviewable document.
![Workflow run details showing a structured The Startup Finance newsletter issue with placed media and applied skill versions]()
Choose `No media` for a copy-only issue, `Use supplied Atomic assets` for reviewed project media, or `Discover and insert media` to let the step reuse project images, generate source-backed charts, and find images or GIFs. Discovered external media is imported into project-scoped Atomic storage before it is embedded. Every inserted asset has a reviewable Atomic URL, alt text, placement, source, and rights status. Reviewers can remove or replace an asset without regenerating the copy.
Run details show the exact base and target skill versions used for the draft. They also provide Markdown and HTML exports rendered from the same structured issue, so the visible document and exports cannot drift.
### Publish an approved post to X
Connect an X account in **Project settings → Integrations**, then add `Publish social post` after `Human review`. X is the only supported social publishing platform in this version. The step publishes the approved `body` for a post or the ordered approved `segments` for a thread exactly as received; it never rewrites or shortens the copy.
This step creates an immediate public external post. Select the intended account, confirm the warning when testing, and keep `Human review` directly before publishing. Run details link every created post. If a thread stops partway through, Atomic preserves the successful post IDs and a retry continues from the first unpublished segment. A timeout whose outcome cannot be proven is marked for reconciliation to avoid blindly creating a duplicate.
The `Repurpose and publish to X` template includes the generation, review, and publication handoff. Scheduling, media uploads, quote posts, editing, deletion, and analytics are not included.
### Human review steps
![Workflow editor Human review step with output variable and output value settings grouped together]()
Use `Human review` when a workflow should pause for a person to approve a value or choose from generated options.
For list reviews, give the output a clear variable name. When the list is an array of JSON objects, choose the preview field reviewers should see in each row. Full JSON remains available in review details. Keep the full selected object as the downstream output unless the next step only needs one nested value.
### Programmatic page workflows
Programmatic workflow blocks are reusable. You can combine only the steps needed for the job:
* `Detect programmatic page clusters`: find repeated competitor URL patterns.
* `Programmatic market research`: combine the description with the current project and brand kit, identify direct competitors and similar projects, inspect how they implement the requested page pattern, and return one typed `pageStructure`. The structure includes the recommended page type, WDF/IDF terms, source URLs, layout requirements, and observed content-length ranges.
* `Extract programmatic page structure`: inspect a `urls` array and return a reusable page generation schema, shared section structure, and full-page screenshot URLs for each analyzed example. Competitor alternative pages include named sections such as hero, trust bar, comparison table, differentiator cards, testimonials, pricing, FAQ, and final CTA.
* `Get Webflow collection structure`: read fields from an existing Webflow CMS collection.
* `Create programmatic collection structure`: create or reuse a Webflow CMS collection from an extracted page structure or existing collection structure.
* `Generate structured page content`: generate one CMS-ready page object from a writing prompt and `pageStructure`. The step runs visible Research and Writing phases, validates required CMS fields and content-length ranges, then returns `generated_page`.
* `Publish mapped content`: map workflow values into WordPress or Webflow and publish or save drafts. For Webflow, use `Map fields` to fill fields one by one, or `Paste JSON` to pass the whole `generated_page` object.
### Map WordPress custom fields
When you select a WordPress connection in `Publish mapped content`, Atomic reads the ACF fields exposed for the selected post or page type. Use `Map fields` to map workflow variables or fixed values one by one. Use `Paste JSON` to provide the complete ACF values object from one workflow variable or a pasted JSON object. Standard WordPress fields such as title, content, excerpt, slug, author, categories, and tags remain separate mappings. Author accepts a WordPress user ID or an exact username, slug, or display name.
Use `Get content from editor` before publishing an existing Atomic content item. It returns one canonical field for each saved value: `title`, `fullDescription`, `metaDescription`, `focusKeyword`, `featuredImageUrl`, `canonicalUrl`, `slug`, `content`, `wordCount`, `readTimeMinutes`, status and score fields, linked record IDs, outline and NLP data, content structure, and timestamps. `fullDescription` preserves the complete summary; `metaDescription` is the short SEO version and is limited to 80 characters, ending in `...` when it must be shortened. Read time is rounded up at 200 words per minute. The slug comes from the canonical URL path when available and otherwise from the title.
![Generate article workflow step showing separate full description and meta description fields]()
![Publish mapped content step showing discovered WordPress ACF fields for Tag and Read Time]()
![Publish mapped content step using Paste JSON for a complete WordPress ACF values object]()
If the custom fields do not appear, open the field group in WordPress and enable `Show in REST API`. Atomic sends the mapped values in the same WordPress publish request as the post or page.
Use `Programmatic market research` when you need to discover what the market expects before creating a collection. Pass the same `pageStructure` into both `Create programmatic collection structure` and `Generate structured page content`. Use `Extract programmatic page structure` when you already know which URLs should define the pattern. Use `Get Webflow collection structure` when you already have a Webflow collection and only need more pages for it.
For an existing Webflow collection, you can optionally select a reference CMS item in `Get Webflow collection structure`. Atomic uses the selected item's field values as bounded examples so generation understands each field's intended role, format, and level of detail. Reference values guide structure only; competitor-specific facts and claims are not copied into the new item.
During `Generate structured page content`, the Research phase searches the selected project knowledge bases first. It then searches the live web and reads the strongest source pages needed to verify competitor and comparison claims. The Writing phase receives that evidence brief and the Webflow schema, but has no research tools; it produces the final CMS field values without inventing unsupported facts. Open Agent view on a completed workflow run to inspect the separate Research and Writing conversations.
Agents can start the same project-aware capability with the `Programmatic Market Research` tool (`programmatic_market_research`). It returns a `researchId` immediately so long-running live research does not depend on one MCP request staying open. Results typically take around 10 minutes, which is expected. Call `wait_for_programmatic_market_research` to wait for completion, or `get_programmatic_market_research` to check the current status and retrieve the canonical `pageStructure`. The workflow step remains self-contained and returns the completed market-research structure directly.
![Workflow editor Programmatic Pages sidebar showing Programmatic market research]()
For Webflow collections, `pageStructure` preserves the exact Webflow type for every field. Short scalar copy should stay `PlainText`. Repeated blocks, formatted lists, tables, embeds, and multi-paragraph content should be `RichText` so collection creation, generation, and publishing all use the same type.
`Generate structured page content` accepts only `prompt` and `pageStructure`. Put the audience, topic, CTA, constraints, and known facts in the prompt. The step fails instead of silently accepting unknown fields, missing required fields, or values outside researched content-length ranges.
The `Webflow programmatic collection structure` template starts with a collection idea and sample-page prompt. It researches the market, creates the matching Webflow collection from the returned typed structure, generates a sample CMS page, and exposes both the research and generated object in the workflow output.
![Workflow output showing the programmatic page blueprint with page setup fields and section cards]()
`Create programmatic collection structure` also returns a Webflow wireframe image. Use `wireframeImageUrl` or `wireframeMarkdown` to review the intended Collection Page layout, including visible hero/body bindings and SEO metadata fields that should be configured in Webflow page settings instead of placed as text blocks on the canvas.
Paste the returned `designerExtensionScript` into the Atomic Webflow app while the Collection Page is open. The app runs that installer directly, so its heading hierarchy and CMS bindings stay identical to the generated script.
![Workflow run output showing a generated programmatic Webflow CMS page and its mapped field values]()
![Workflow editor Publish mapped content step with Paste JSON mapping mode selected]()
When a generated page already matches the Webflow collection fields, select `Paste JSON` and use `{{generateStructuredPageContent.generated_page}}`.
## Workflow run tab (inputs and output)
![Workflow detail Run tab showing inputs and output panel]()
This is the control panel for a specific workflow:
* Set run inputs (for example, `Current Date`).
* Confirm brand context before run.
* Click `Run workflow` to execute.
* Read final output in the right panel.
Use this view when you need a clean rerun with controlled inputs.
## Execution tab (live step progress)
![Workflow detail Execution tab showing live step progress]()
Execution helps you debug in real time:
* Left side shows each step and status.
* Right side shows details for the selected step.
* Running steps reveal where time is spent.
* Click `Cancel run` to stop the selected active run without canceling the workflow itself.
Use this tab to find bottlenecks, stalled steps, or weak step prompts.
## Run progress state
![Workflow detail showing a step currently running]()
During a live run, watch for:
* Step stuck in running too long.
* Empty or partial step output.
* Unexpected transitions between steps.
If one step repeatedly slows or fails, simplify that step before scaling schedule.
## Workflow history tab (single workflow)
![Workflow detail History tab with run status and AI credit usage]()
Use this tab to review reliability over time for one workflow:
* Compare run outcomes.
* Confirm stopped runs show `canceled`.
* Track AI credit trend.
* Validate cadence against business needs.
* Re-open older runs for QA checks.
This is the best view for per-workflow quality review.
## Schedule a workflow (command action)
![Schedule workflow dialog opened from workflow detail actions]()
From workflow detail actions, open `Schedule workflow` and define cadence:
* Daily
* Weekly
* Bi-weekly
* Monthly
* Inactive
Set schedule only after the workflow is stable in manual runs.
## What to automate first
Start with tasks that are frequent and rules-based:
* Daily or weekly priorities for SEO and content.
* Content refresh candidate detection.
* Alerting for major movement or failures.
* Structured summaries for stakeholders.
Keep one-off strategy prompts in New chat.
## What to fix first
If runs are unstable, fix in this order:
1. Inputs (wrong or missing context)
2. Step prompt clarity
3. Step order and dependencies
4. Schedule frequency
Most failures come from bad inputs or vague step instructions.
Important: Completed status does not guarantee a useful
output. Always review output quality before scheduling at scale.
## Weekly workflow checklist
1. Review failed runs.
2. Check output quality for completed runs.
3. Improve one weak step in the top-used workflow.
4. Remove duplicate workflows and keep one source of truth.
5. Reconfirm schedule only for proven workflows.
## Keep in mind
* Small reliable workflows beat large fragile workflows.
* Naming matters for discoverability and adoption.
* A successful status does not always mean useful output.
* Schedule is an operations tool, not a quality fix.
## Where to go next
* [Agents](/automation/agents/overview): assign the right specialist to each step
* [Automations](/automation/automations/overview): schedule recurring agent conversations around your workflows
* [Tasks](/automation/tasks/overview): track follow-through after workflow output
* [New chat](/automation/new-chat/overview): prototype tasks before turning them into workflows
# Citations
Source: https://docs.atomicagi.com/data/ai-search/citations
Understand who gets cited in AI answers and where your brand needs stronger source authority
![AI Search Citations page with trend and By page table]()
Use this page to improve the source quality behind your AI visibility.
Important: Citation quality matters more than citation
volume. Prioritize trusted domains that influence decisions.
## Questions this page should answer
1. Which domains are getting citation share in our topic space?
2. Is our brand cited enough on strategic prompts?
3. Where should we build authority first?
## Before you analyze
* Keep date range aligned with Visibility and Competitors.
* Start with all prompts and all platforms.
* Then segment by citation type and platform.
## What this page gives you
* Citation-share trend across top cited domains.
* Ranked citation table with prompt context.
* Two subtabs: `By page` and `By domain`.
## How to read the top citation trend
* `Avg. share`: average citation footprint.
* `Rank`: relative placement in citation results.
* `Type`: source class (`You`, `Competitor`, `Editorial`, `UGC`, and others).
Key signal: If competitor citations rise on strategic
prompts while your share stays flat, authority is shifting against your
brand.
Example: You can keep visibility on a prompt while losing
citation share. That usually means you are still mentioned, but trusted
sources are increasingly backing competitors.
## How these metrics are calculated (simple)
### Avg. share
```text theme={null}
Avg. share = (Citations for a domain or URL / total citations in the selected scope) x 100
```
### Rank
Rank is the ordering after sorting by `Avg. share` (`rank 1` has the highest share).
### Type
Type is a rule-based source classification (`You`, `Competitor`, `Editorial`, `UGC`, and others).
## By page tab
Use this tab for URL-level citation diagnosis.
![AI Search Citations with By page tab selected]()
Prioritize pages where:
* Competitors are repeatedly cited on important prompts.
* Your page appears but with low share.
* Citation type is weak for a high-value query cluster.
## By domain tab
Use this tab to evaluate domain-level authority distribution.
![AI Search Citations with By domain tab selected]()
Use it to:
* Separate durable editorial authority from lower-signal sources.
* Identify domains where your brand is underrepresented.
* Plan outreach and evidence content by domain priority.
## Quick weekly checklist
1. Track top citation winners and losers.
2. Compare your brand share by page and domain.
3. Flag prompt clusters with weak source support.
4. Assign one source-authority action.
## How to use filters
* `All prompts`: isolate strategic prompt groups.
* `All platforms`: compare source behavior across engines.
* `All types`: focus on source class (`Editorial`, `UGC`, `You`, and others).
## What to fix first
| Pattern in Citations data | What it usually means | Recommended action |
| ------------------------------------ | --------------------------- | ------------------------------------------- |
| Competitor citations rising | Trust gap widening | Publish stronger proof-led content |
| Your domain appears with low share | Citation weight is weak | Improve evidence quality and source context |
| UGC share high, editorial share low | Authority profile imbalance | Build editorial/institutional mentions |
| Visibility stable, citations falling | Position is less defensible | Strengthen source-quality foundations |
## Team routine
1. Weekly: monitor citation share changes.
2. Bi-weekly: review source-type quality mix.
3. Monthly: report citation progress on strategic prompts.
## Keep in mind
* Citation quality matters more than raw citation count.
* Authority changes often lag content publication.
* Domain-level wins are usually more durable than single-page wins.
## Where to go next
* [Visibility](/data/ai-search/visibility)
* [Competitors](/data/ai-search/competitors)
* [Prompts](/data/ai-search/prompts)
* [Overview](/data/ai-search/overview)
# Competitors
Source: https://docs.atomicagi.com/data/ai-search/competitors
Benchmark your brand against competitors in AI answers and decide where to recover share first
![AI Search Competitors page with trend chart and competitor benchmark table]()
Use this page to compare your performance against direct competitors in one place.
Important: Compare the same prompt set and date range
before declaring a competitor win/loss.
## Questions this page should answer
1. Which competitors are strongest on our tracked prompts?
2. Where are we losing on visibility, position, or citations?
3. Which competitive gap is most recoverable this sprint?
## Before you analyze
* Keep the same date range as Visibility.
* Start with broad prompt and platform scope.
* Focus on repeated movement, not one-day spikes.
## What this page gives you
* Trend comparison between your brand and competitors.
* Table with `Visibility %`, `AVG position`, and `Citations`.
* Fast prioritization of threats and opportunities.
## How to read this page correctly
* `Visibility %`: share of appearance in tracked AI answers.
* `AVG position`: average placement quality.
* `Citations`: source-backed presence strength.
Key signal: A competitor with lower visibility but stronger
citations can still become a near-term threat because trust signals are
improving faster.
## How these metrics are calculated (simple)
### Visibility %
```text theme={null}
Visibility % = (Responses where brand appears / total evaluated responses) x 100
```
### AVG position
```text theme={null}
AVG position = Sum of answer placement indexes / Number of responses where that brand appears
```
Lower value is better.
### Citations
Citations are the total source references attributed to that brand/domain in evaluated responses.
## Quick weekly checklist
1. Compare your brand with top 3 competitors.
2. Flag one competitor gaining fastest.
3. Identify one gap you can close in 1-2 sprints.
4. Assign actions to prompts and pages.
## How to use filters
* Prompt scope: isolate the commercial topic you care about.
* Platform scope: find engine-specific competitive pressure.
* Date range: validate whether movement is persistent.
## What to fix first
| Pattern in Competitors data | What it usually means | Recommended action |
| ------------------------------------ | ----------------------- | -------------------------------------------------- |
| Competitor visibility rising | Better answer relevance | Improve topical depth and prompt alignment |
| Competitor citations much stronger | Better trust footprint | Increase authority sources and proof quality |
| You lead visibility, lag on outcomes | Post-click weakness | Improve destination pages and conversion clarity |
| Losses concentrated on one platform | Platform-specific gap | Adapt content structure for that platform behavior |
## Team routine
1. Weekly: review top movers.
2. Bi-weekly: run one targeted recovery sprint.
3. Monthly: report share movement versus top competitors.
## Keep in mind
* Competitive sets evolve with prompt mix.
* Share gains without citation support are fragile.
* Short reaction cycles outperform quarterly-only reviews.
## Where to go next
* [Visibility](/data/ai-search/visibility)
* [Citations](/data/ai-search/citations)
* [Prompts](/data/ai-search/prompts)
* [Pages](/data/ai-search/pages)
# Generative engines
Source: https://docs.atomicagi.com/data/ai-search/generative-engines
Compare AI engines side by side to decide where to focus optimization this week
![AI Search Generative engines page with trend chart and engine performance table]()
Use this page to decide which engine deserves immediate work and which engine can stay in maintenance mode.
Important: Rank engines by business value first
(Conversions), not by traffic volume alone.
## Questions this page should answer
1. Which engine currently drives the highest-value traffic?
2. Which engine is losing momentum fastest?
3. Where should we prioritize prompt and page updates this sprint?
## Before you analyze
* Use the same date range as Overview.
* Compare outcomes, not only traffic.
* Check at least two periods before making strategic changes.
## What this page gives you
* Engine-level trend lines for key platforms.
* A table with `AI clicks`, `Conversions`, and `Avg. time` by engine.
* A fast way to separate platform issues from content issues.
## How to read this page correctly
* `AI clicks` tells you volume.
* `Conversions` tells you business value.
* `Avg. time` is a quality proxy for post-click engagement.
Common patterns:
* High clicks, weak conversions: intent mismatch.
* Lower clicks, stronger conversions: high-value traffic.
* Drop on one engine only: platform-specific fit issue.
Key signal: Conversion leaders should drive your roadmap.
Treat high-click, low-conversion engines as optimization targets, not
winners.
## How these metrics are calculated (simple)
### AI clicks
AI clicks are the number of AI-attributed sessions grouped by platform.
### Conversions
Conversions are the selected GA4 conversion events generated by those sessions.
### Avg. time
```text theme={null}
Avg. time = Total engaged time from AI-attributed sessions / Number of AI-attributed sessions
```
## Quick weekly checklist
1. Rank engines by conversions first.
2. Flag the biggest week-over-week loser.
3. Assign one action per weak engine.
4. Recheck impact in the next reporting cycle.
## How to use filters
* Keep date range aligned with Overview.
* Compare one engine hypothesis at a time.
* Avoid changing multiple filters while diagnosing one issue.
## What to fix first
| Pattern by engine | What it usually means | Recommended action |
| ------------------------------------ | -------------------------------- | -------------------------------------------------- |
| Clicks and conversions both falling | Visibility and relevance decline | Improve prompt coverage and content fit |
| Clicks up, conversions down | Low-intent traffic increase | Improve query intent mapping and destination pages |
| Avg. time falls while clicks stay up | Weak post-click alignment | Improve page structure and value framing |
| One engine strongly outperforms all | Channel concentration risk | Replicate winning approach on weaker engines |
## Team routine
1. Weekly: review engine winners and losers.
2. Bi-weekly: compare conversion efficiency by engine.
3. Monthly: rebalance effort by engine ROI.
## Keep in mind
* Engine behavior can shift after model updates.
* Volume leadership is not always value leadership.
* Platform fit changes faster than quarterly planning cycles.
## Where to go next
* [Prompts](/data/ai-search/prompts)
* [Pages](/data/ai-search/pages)
* [Competitors](/data/ai-search/competitors)
* [Overview](/data/ai-search/overview)
# Mentions
Source: https://docs.atomicagi.com/data/ai-search/mentions
Separate existing brand mentions from missing backlink opportunities so you can focus authority work where it matters
![AI Search Mentions page with mention and backlink opportunity data]()
Use this page to review external AI citation sources through a mentions lens. It helps you separate sources that already mention your brand from sources that still represent an authority gap.
## Questions this page should answer
1. Which sources already mention our brand?
2. Where are we missing backlinks or citation coverage?
3. Which mention gaps should we prioritize first?
## Before you analyze
* Keep the same date range as Citations and Visibility.
* Start with all platforms before filtering to one engine.
* Use prompt and topic filters when checking a specific narrative gap.
## What this page gives you
* Mention and non-mention source rows from AI citation data.
* Filters for platform, prompt, topic, and source type.
* A direct way to spot existing brand coverage versus missing opportunities.
## How to read mention rows
Read each row as a source opportunity, not just a backlink lead.
* `Source`: the domain or page appearing in AI citation context.
* `Mention status`: whether your brand is already mentioned.
* `Prompt/topic`: where the source appeared.
* `Platform`: the AI engine context where the source was observed.
* `Source type`: the kind of site or content you may need to influence.
Use this rule:
* Existing mentions reveal proof patterns worth repeating.
* Missing mentions on high-value prompts deserve outreach or stronger source-backed content.
* Missing mentions on low-value prompts should not distract from strategic authority work.
## How these metrics are calculated (simple)
```text theme={null}
Mention = cited source includes or references your brand
```
```text theme={null}
Missing mention = cited source appears in relevant AI context but does not mention your brand
```
Use this page with [Citations](/data/ai-search/citations) to understand both source presence and brand inclusion.
## How to use this page
### Prioritize strategic missing mentions
Missing mentions on high-value prompts matter more than isolated misses on low-value topics. Focus on gaps tied to commercial or high-trust topics first.
### Check the source type before acting
Editorial, institutional, reference, and corporate sources usually deserve a different outreach or content strategy than UGC sources.
### Use mention wins as proof patterns
When a source already mentions you, inspect what kind of page or evidence got picked up. That pattern often tells you what to expand next.
## Quick weekly checklist
1. Filter to your highest-priority prompt or topic.
2. Flag sources where competitors are present and you are missing.
3. Flag strong mention wins worth replicating.
4. Turn the top 3 gaps into outreach or content proof actions.
## What to fix first
| Pattern in Mentions data | What it usually means | Recommended action |
| ------------------------------------- | -------------------------------------------------- | -------------------------------------------------------- |
| Missing mentions on buying prompts | Authority gap affects revenue intent | Build proof, outreach, or comparison assets |
| Strong sources mention you | Existing proof pattern works | Replicate the source/content pattern |
| Many UGC misses | Not all gaps are worth pursuing | Prioritize editorial, institutional, or category sources |
| Competitors appear repeatedly | They have stronger external proof | Review competitor citations and source strategy |
| Mentions exist but visibility is weak | Source coverage is not translating to answer share | Pair with Visibility and Prompts diagnostics |
## Team routine
1. Weekly: review high-value missing mentions.
2. Bi-weekly: turn the best gaps into outreach or proof-building work.
3. Monthly: compare mention wins with citation share and visibility changes.
## Keep in mind
* Not every missing mention is worth chasing.
* A strong mention on one source does not guarantee broader authority.
* Use this page with Citations, not instead of Citations.
## Where to go next
* [Citations](/data/ai-search/citations)
* [Visibility](/data/ai-search/visibility)
* [Competitors](/data/ai-search/competitors)
* [Competitors overview](/data/competitors/overview)
# AI Search Overview
Source: https://docs.atomicagi.com/data/ai-search/overview
Start here to understand AI traffic direction, source quality, and which pages need action first
![AI Search Overview top section with AI clicks and total conversions trend]()
Use this page as your weekly control center for AI Search. It tells you where performance is moving before you jump into deeper pages.
Important: Review this page first, then drill down. It
prevents teams from optimizing the wrong source or page.
## Questions this page should answer
1. Are AI clicks and conversions moving in the same direction?
2. Which AI sources are helping most, and which are weakening?
3. Which pages should we prioritize first this sprint?
## Before you analyze
* Set the same date range you will use on the other AI Search pages.
* Compare against an equivalent prior period.
* Start unfiltered so you see the global pattern first.
## What this page gives you
* A top trend for `AI clicks` and `Total conversions`.
* A `By source` table to compare AI engines.
* A `By page` table to see which URLs drive AI traffic quality.
## How to read the top trend correctly
* `AI clicks`: sessions coming from AI platforms.
* `Total conversions`: tracked outcomes after those visits.
Read both together:
* Clicks up, conversions down: traffic quality or intent fit is weak.
* Clicks flat, conversions up: landing page quality likely improved.
* Clicks down, conversions down: visibility or prompt coverage problem is likely.
Key signal: If clicks and conversions both decline, start
with visibility and prompt coverage checks before page-level edits.
Example: If ChatGPT clicks drop 22% and conversions drop
18% in the same week, first review Visibility and Prompts for that source
before changing landing pages.
## How these metrics are calculated (simple)
### AI clicks
AI clicks are the total sessions attributed to supported AI referrers in the selected date range.
### Total conversions
Total conversions are the selected GA4 conversion events generated by those AI-attributed sessions.
## By source section
![AI Search Overview By source and By page tables]()
Use `By source` to decide engine focus for the week:
* Prioritize sources with strong conversion efficiency.
* Flag sources with sharp drop in clicks and conversions.
* Avoid over-investing in sources that drive volume but no business impact.
## By page section
In the same table view, use `By page` to choose URL-level actions:
* Pages with high AI clicks and weak avg. time need better on-page clarity.
* Pages with stable traffic but falling outcomes need message refinement.
* Pages with improving outcomes are templates for replication.
## Quick weekly checklist
1. Confirm top-line direction from clicks and conversions.
2. Pick one growth source and one declining source.
3. Pick three URLs for immediate improvement.
4. Route deeper analysis to the right page: Visibility, Prompts, Pages, or Citations.
## How to use filters
* Use date range first to keep reporting consistent.
* Narrow by one segment at a time when validating a hypothesis.
* Do not compare mixed filter sets in the same decision.
## What to fix first
| Pattern in Overview data | What it usually means | Recommended action |
| -------------------------------- | ------------------------------------------- | ------------------------------------------------ |
| Clicks down and conversions down | Visibility or prompt relevance loss | Check `Visibility` and `Prompts` first |
| Clicks up, conversions flat | Traffic quality mismatch | Improve destination pages and intent alignment |
| One source drops sharply | Platform-specific issue | Investigate `Generative engines` and competitors |
| Stable traffic, weaker avg. time | Content experience mismatch after the click | Improve page structure and value clarity |
## Team routine
1. Weekly: run Overview first before any deep dive.
2. Bi-weekly: compare source efficiency changes.
3. Monthly: report AI Search contribution to business outcomes.
## Keep in mind
* AI traffic can be volatile in short windows.
* Conversion lag can hide recent improvements.
* Top-line trend is useful only when paired with source and page context.
## Where to go next
* [Generative engines](/data/ai-search/generative-engines)
* [Visibility](/data/ai-search/visibility)
* [Pages](/data/ai-search/pages)
* [Prompts](/data/ai-search/prompts)
* [Citations](/data/ai-search/citations)
* [Sentiment](/data/ai-search/sentiment)
# Pages
Source: https://docs.atomicagi.com/data/ai-search/pages
Prioritize URLs that drive AI visits, recover zero-click pages, and improve indexability quality
![AI Search Pages with Landing pages tab selected, showing top cards, trend chart, and table]()
Use this page as your URL-level execution dashboard for AI Search.
Important: Treat this as a priority queue, not just a
report. Pick pages, assign owners, and track movement next cycle.
## Questions this page should answer
1. Which pages are currently driving AI visits?
2. Which pages are visible but not getting clicks?
3. Which URLs should we update first this sprint?
## Before you analyze
* Use the same date range as Overview.
* Start with no extra filters.
* Compare against an equivalent prior period.
## What this page gives you
* Top cards for `Landing pages`, `Zero-click pages`, and `AI indexability percentage`.
* A trend + table view to prioritize by URL.
* Two subtabs for execution: `Landing pages` and `Zero-click pages`.
## How to read the top cards
* `Landing pages`: URLs currently receiving AI traffic.
* `Zero-click pages`: URLs mentioned but not clicked.
* `AI indexability percentage`: share of pages discoverable for AI contexts.
Key signal: If zero-click pages increase while landing-page
outcomes stall, your visibility is not translating into qualified traffic.
## How these metrics are calculated (simple)
### Landing pages
Landing pages are the count of unique URLs with at least one AI-attributed session in the selected period.
### Zero-click pages
Zero-click pages are tracked URLs mentioned in AI answers but with zero AI-attributed sessions.
### AI indexability percentage
```text theme={null}
AI indexability percentage = (AI-discoverable tracked URLs / total tracked URLs) x 100
```
## Landing pages tab
Use this tab for pages already receiving traffic and needing performance gains.
![AI Search Pages Landing pages tab with trend and by-page performance table]()
Focus on:
* High clicks, low avg. time pages.
* Former winners now trending down.
* Pages with potential to improve conversion quality.
## Zero-click pages tab
Use this tab to convert mention visibility into actual traffic.
![AI Search Pages Zero-click pages tab with zero-traffic URL table]()
Focus on:
* High-value URLs repeatedly shown with no clicks.
* Pages with weak titles or weak value framing.
* Pages needing better internal paths from strong pages.
## Quick weekly checklist
1. Pick 3 landing pages to optimize now.
2. Pick 3 zero-click pages to recover.
3. Assign clear owner and deadline per URL.
4. Check movement next cycle.
## How to use filters
* Start broad, then narrow to one topic cluster.
* Keep date range consistent across Pages and Overview.
* Validate changes in one segment before scaling.
## What to fix first
| Pattern in page data | What it usually means | Recommended action |
| ----------------------------------- | -------------------------------------- | --------------------------------------------------- |
| High visibility, no clicks | Weak click motivation or mismatch | Improve page positioning and relevance |
| Clicks dropping on core URL | Content is stale or less competitive | Refresh content depth and proof |
| Strong clicks, weak avg. time | Post-click expectation mismatch | Improve structure and immediate value communication |
| Many zero-click URLs in one cluster | Topic is visible but weakly compelling | Strengthen topic authority and differentiation |
## Team routine
1. Weekly: ship page updates on highest-value URLs.
2. Bi-weekly: compare cohort performance after changes.
3. Monthly: rebalance roadmap by URL-level outcomes.
## Keep in mind
* Not every zero-click URL is urgent.
* Avg. time is directional, not absolute.
* URL improvements can lag one cycle behind prompt changes.
## Where to go next
* [Prompts](/data/ai-search/prompts)
* [Visibility](/data/ai-search/visibility)
* [Citations](/data/ai-search/citations)
* [Overview](/data/ai-search/overview)
# Prompts
Source: https://docs.atomicagi.com/data/ai-search/prompts
Manage tracked prompts, identify ranking gaps, and prioritize the questions that need optimization first
![AI Search Prompts page with top cards and prompt-level performance table]()
Use this page to decide whether your tracked prompt set matches real buyer demand.
Important: Better prompt coverage beats higher prompt
count. Focus on business-critical prompts first.
## Questions this page should answer
1. Are we tracking the prompts that matter most for pipeline?
2. Which prompts are underperforming right now?
3. Which prompt gaps should become this sprint's content work?
## Before you analyze
* Review the active prompt list before judging performance.
* Keep the same reporting window used in other AI Search pages.
* Compare by intent category, not only by total rows.
## What this page gives you
* Top prompt health indicators like average visibility and average position.
* Prompt-level table with position, mention frequency, top results, and category.
* A `Tracking engines` control for choosing which plan-supported engines run each active prompt.
* `Manage prompts` flow with one add workspace that combines manual entry and AI-generated suggestions.
* A decision layer to connect prompt gaps to page updates.
## How prompt data is collected
Atomic runs each active prompt on the engines selected in `Tracking engines`. Available engines and automation frequency depend on
your plan. We then compare patterns over time, because AI responses naturally vary between runs.
These tracked prompts should be written as conversational questions (for example `What's the best CRM for marketing agencies under
50 people?`), not short keyword fragments.
To mirror real buyer behavior, Atomic collects prompt results through AI product interfaces, not only via API outputs. This helps
you measure what users are likely to see in real interactions.
Why this matters:
* `Authentic experience`: data reflects the same interface-level outputs users interact with.
* `Real-world accuracy`: results are less abstract than API-only snapshots.
* `Broader coverage`: you can track platforms and model flows with limited public API access.
## Choose tracking engines
Click `Tracking engines` in the top-right corner of the Prompts page. Select the engines available in your plan, then save the
configuration. Starter projects can select up to three engines from ChatGPT, Perplexity, AI Mode, and Gemini.
![Tracking engines dialog showing selected AI search engines and calculated prompt capacity]()
Your plan provides prompt-run capacity rather than a fixed number of distinct prompts:
```text theme={null}
Used prompt runs = active prompts x selected engines
Maximum active prompts = plan prompt runs / selected engines
```
For example, the Starter plan's 75 prompt runs can cover 75 prompts on one engine, 37 prompts on two engines, or 25 prompts
on three engines.
Selecting fewer engines increases distinct-prompt capacity. Historical results remain available after an engine is deselected;
only future runs use the new selection.
If a plan change makes the current selection exceed capacity, tracking pauses without deleting prompts or engine choices. Open
`Tracking engines` and reduce the engine count, or deactivate prompts, until the configuration fits.
## How to read the top cards
* `Avg. visibility percentage`: average inclusion across tracked prompts.
* `Your avg position`: average placement quality.
Use both:
* Better position with low visibility means narrow coverage.
* Visibility up with weak position means presence without competitiveness.
Key signal: Prioritize prompts where mention frequency is
high but your average position is weak. Those are usually your fastest
visibility gains.
## How these metrics are calculated (simple)
### Avg. visibility percentage
```text theme={null}
Avg. visibility percentage = (Number of tracked prompts where your brand is present / total tracked prompts) x 100
```
### Your avg position
```text theme={null}
Your avg position = Sum of your placement indexes / Number of prompt responses where your brand appears
```
Lower value is better (closer to top placement).
### Mention frequency
```text theme={null}
Mention frequency = (Responses mentioning your brand for that prompt / Total sampled responses for that prompt) x 100
```
## How to read the prompt table
* `Your position`: current placement for that prompt.
* `Mention frequency`: how often your brand appears.
* `Top results`: recurring sources shown in answers.
* `Category`: funnel context.
Prioritize prompts that are high intent, high frequency, and weak in position.
## How to add new prompts
Use the top-right `Manage prompts` button on this page, then click `Add prompt`.
The left sidebar shows current groups and how many prompt slots are already used for the selected engine count.
Inside `Add prompt`, the current UI keeps both actions in one workspace:
1. Add one prompt manually with stage and group controls at the top.
2. Review AI-generated suggestions in the list below and add the ones you want.
### Add prompt workspace
![Manage prompts add workspace with manual prompt fields at the top and suggested prompt list below]()
Use this workspace when you are updating the tracked prompt set, not only when you are adding suggestions.
At the top of the panel you can:
* Enter one prompt manually.
* Set the funnel stage.
* Let Atomic auto-detect the group or choose one yourself.
* Click `Add prompt` to save the new tracked prompt.
Below that, Atomic shows suggested prompts with group, stage, and confidence.
How suggestions are generated:
* Atomic analyzes your project domain plus top AI search pages from the last 30 days.
* The system generates prompt ideas across funnel categories (`awareness`, `consideration`, `decision`, `branded`).
* Each suggestion includes a confidence score to help you prioritize.
How to use suggested prompts:
1. Click `Manage prompts`.
2. Click `Add prompt`.
3. Review prompt text, group, stage, and confidence in the `Suggested` list.
4. Use `Generate suggestions` when you want a fresh batch of ideas.
5. Add the prompts that best match current pipeline goals.
### Manual entry
![Manage prompts add workspace showing manual prompt textarea, stage selector, and Add prompt button]()
Use manual entry when you already know the exact prompt to track (for example from sales calls, campaign briefs, or competitor monitoring).
How to add manually:
1. Click `Manage prompts`.
2. Click `Add prompt`.
3. Enter one prompt in natural language.
4. Set the stage and confirm or change the group.
5. Click `Add prompt`.
## Quick weekly checklist
1. Review decision-intent prompts first.
2. Remove low-signal prompts that do not map to business goals.
3. Add missing high-intent prompts from sales conversations.
4. Assign one action per priority prompt.
## How to use filters
* Use category filters to separate awareness and decision queries.
* Use date range to confirm trend stability.
* Change one filter at a time when diagnosing.
## What to fix first
| Pattern in prompt data | What it usually means | Recommended action |
| --------------------------------------- | -------------------------- | ---------------------------------------------- |
| Decision prompts rank poorly | Commercial relevance gap | Improve service-page fit and proof depth |
| Awareness strong, decision weak | Funnel imbalance | Add decision-stage prompt coverage |
| Mention frequency low across categories | Authority and coverage gap | Improve citations and topical breadth |
| Good positions, weak traffic impact | Prompt set too narrow | Expand strategically similar prompt variations |
## Team routine
1. Weekly: refresh priority prompt backlog.
2. Bi-weekly: audit category balance.
3. Monthly: align prompt coverage with pipeline goals.
## Keep in mind
* Prompt quality is more important than prompt count.
* One viral query can distort short windows.
* Repeated gaps are more important than one-off misses.
## Where to go next
* [Pages](/data/ai-search/pages)
* [Competitors](/data/ai-search/competitors)
* [Citations](/data/ai-search/citations)
* [Overview](/data/ai-search/overview)
# Questions
Source: https://docs.atomicagi.com/data/ai-search/questions
Review question-shaped queries so you can find informational demand and decide which answers need better coverage
![AI Search Questions page with the Questions tab selected]()
Use this page to review question-style search demand in one place. It helps you decide which educational or comparison questions deserve new content, better answers, or clearer page targeting.
## Questions this page should answer
1. Which questions are already bringing visits?
2. Which question patterns are rising or falling?
3. Which pages should answer those questions better?
## Before you analyze
* Keep the date range aligned with your other search reports.
* Start with broad question themes before drilling into single queries.
* Compare performance with landing pages before changing content.
## What this page gives you
* A table of question-style queries.
* Click, impression, CTR, and position context for those questions.
* Direct drill-down from a question to the related landing pages.
## How to read the top question trend
Use the chart to spot movement before opening individual rows.
* The selected question chips show the questions currently plotted.
* `Clicks` shows whether the question is already producing visits.
* `Impressions` shows whether demand exists even when clicks are weak.
* `Position` shows whether the page is close enough to improve with focused work.
* `CTR` shows whether the current result earns attention.
Use this rule:
* High impressions + low CTR usually means the answer snippet or page framing is weak.
* Rising clicks on repeated themes can justify a new section, FAQ, or comparison page.
* Poor position on high-intent questions usually needs stronger page relevance and links.
## How these metrics are calculated (simple)
```text theme={null}
Question row = search query classified as question-style intent
```
```text theme={null}
CTR = clicks / impressions
```
Position is the average ranking position for that query during the selected date range.
## How to use this page
### Start with repeated themes
Group similar questions together first. Repeated wording around pricing, alternatives, setup, or trust usually points to a missing section or missing page, not just one weak keyword.
### Separate education from decision intent
Some questions need early-funnel explainer content. Others need stronger product pages, comparison pages, or FAQ sections on existing money pages.
### Validate with page-level context
If a question has impressions but weak clicks, check whether the matching page title and description answer the question clearly enough.
## Quick weekly checklist
1. Flag the top rising question cluster.
2. Flag the top declining question cluster.
3. Check which landing pages are responsible for those movements.
4. Turn one high-intent cluster into a content or refresh task.
## What to fix first
| Pattern in Questions data | What it usually means | Recommended action |
| --------------------------- | ------------------------------------------- | ----------------------------------------------------- |
| High impressions, low CTR | Result does not answer the question clearly | Rewrite title, intro, or FAQ answer |
| Good clicks, poor position | Demand exists but page authority is weak | Improve relevance and add internal links |
| Many similar questions rise | Searchers need a dedicated answer pattern | Add a FAQ section or create a focused page |
| Question has buying intent | Commercial page may need clearer comparison | Add proof, pricing, alternatives, or decision content |
| One-off question spikes | Could be temporary noise | Watch before assigning major content work |
## Team routine
1. Weekly: group repeated question patterns.
2. Bi-weekly: refresh the highest-intent page tied to question demand.
3. Monthly: compare question movement with landing page and prompt performance.
## Keep in mind
* A single question rarely matters by itself. Patterns matter more.
* Question demand often reveals missing comparison, FAQ, or trust content.
* Do not rewrite pages until the same question theme appears often enough to matter.
## Where to go next
* [Google Search overview](/data/google-search/overview)
* [Keywords](/data/google-search/keywords)
* [Landing pages](/data/google-search/landing-pages)
* [Prompts](/data/ai-search/prompts)
# Sentiment
Source: https://docs.atomicagi.com/data/ai-search/sentiment
Monitor how AI responses describe your brand and fix negative narrative trends early
![AI Search Sentiment page with trend, positive and negative split, and All tab]()
Use this page to track brand perception quality in AI answers.
Important: One negative mention is noise. Repeated negative
themes across prompts are the real risk.
## Questions this page should answer
1. Is sentiment improving or getting riskier?
2. Which themes are driving positive or negative perception?
3. What messaging updates should we ship first?
## Before you analyze
* Match date range with the rest of AI Search reporting.
* Read trend direction before row-level details.
* Separate positive and negative themes before acting.
## What this page gives you
* Positive sentiment trend line.
* Positive vs negative share summary.
* Theme-level evidence table with sources.
* Three subtabs: `All`, `Positive`, and `Negative`.
## How to read the top sentiment section
* Trend line shows narrative direction over time.
* Right-side split shows current narrative balance.
* Theme rows show what is shaping that balance.
Key signal: If visibility stays stable but negative share
rises, conversion risk is usually increasing before traffic metrics react.
## How these metrics are calculated (simple)
### Positive share
```text theme={null}
Positive share = (Positive-labeled responses / total sentiment-labeled responses) x 100
```
### Negative share
```text theme={null}
Negative share = (Negative-labeled responses / total sentiment-labeled responses) x 100
```
### Theme rows
Theme rows are recurring sentiment statements grouped into actionable theme clusters.
## All tab
Start in `All` for a full narrative baseline.
![AI Search Sentiment with All tab selected]()
Use it to identify:
* Mixed themes requiring clearer positioning.
* Repeated concerns that hurt trust.
* Positive messages worth scaling.
## Positive tab
Use `Positive` to preserve and scale what already works.
![AI Search Sentiment with Positive tab selected]()
Focus on:
* Positive themes that repeat across sources.
* Signals you can reuse in landing pages and prompts.
* Source types that produce high-trust mentions.
## Negative tab
Use `Negative` to reduce risk quickly.
![AI Search Sentiment with Negative tab selected]()
Focus on:
* Recurring negative themes.
* Negative themes tied to decision queries.
* Source patterns that repeatedly create risk.
## Quick weekly checklist
1. Check positive vs negative balance.
2. Flag top recurring negative theme.
3. Protect one high-impact positive theme.
4. Assign one message/proof fix per sprint.
## How to use filters
* Start with `All` before narrowing.
* Compare `Positive` and `Negative` separately.
* Keep window consistent to avoid false trend shifts.
## What to fix first
| Pattern in Sentiment data | What it usually means | Recommended action |
| -------------------------------------- | ---------------------------------- | --------------------------------------------- |
| Negative share rising on trust topics | Credibility concern | Add stronger proof and clearer claims |
| Positive stable, conversions weakening | Narrative not supporting decisions | Improve decision-stage messaging |
| One negative theme repeats | Structural positioning gap | Update core narrative and FAQ/support content |
| Positive themes come from few sources | Fragile narrative base | Expand high-quality source distribution |
## Team routine
1. Weekly: review drift and assign fixes.
2. Bi-weekly: verify whether negative themes were reduced.
3. Monthly: align messaging roadmap with sentiment trends.
## Keep in mind
* Sentiment is directional, not absolute truth.
* Short windows can be noisy.
* Repeated themes matter more than isolated mentions.
## Where to go next
* [Citations](/data/ai-search/citations)
* [Prompts](/data/ai-search/prompts)
* [Pages](/data/ai-search/pages)
* [Overview](/data/ai-search/overview)
# Visibility
Source: https://docs.atomicagi.com/data/ai-search/visibility
Track where your brand is visible across AI platforms and diagnose why share is moving
![AI Search Visibility top section with platform trend and competitor comparison]()
Use this page to monitor brand exposure before click and conversion metrics react.
Important: Visibility is an early signal. If visibility
drops now, clicks and conversions often drop next.
## Questions this page should answer
1. Is our visibility rising or falling by platform?
2. Which competitors are taking share on our tracked prompts?
3. Is the issue in rank quality, citation strength, or recent run outcomes?
## Before you analyze
* Keep date range aligned with Overview.
* Start with `All prompts` and `All categories`.
* Then narrow one segment at a time.
## What this page gives you
* Top visibility trend by platform.
* Competitor performance matrix.
* Four deep sections in this same page: `Positions`, `Citation share`, `Recent chats`, and `Chat detail`.
## Top visibility section
![AI Search Visibility with platform percentages and competitor table]()
Read this section first:
* Platform percentages show who currently owns answer visibility.
* Trend lines show direction and volatility.
* Competitor rows show who is gaining where.
Key signal: If visibility declines before clicks decline,
treat it as an early warning and act before traffic impact appears.
## How these metrics are calculated (simple)
### Visibility %
```text theme={null}
Visibility % = (Responses where brand appears / total evaluated responses) x 100
```
### Avg position
```text theme={null}
Avg position = Sum of rank slots where the brand appears / Number of responses where the brand appears
```
Lower value is better.
### Citation share
```text theme={null}
Citation share = (Citations pointing to the brand or its domains / total citations in scope) x 100
```
## Positions section
Use this section to judge ranking quality, not just visibility share.
![AI Search Visibility Positions section with average position chart and platform table]()
How to read it:
* Lower average position is better.
* Stable visibility with worse position means fragile performance.
* Position deterioration is often an early warning signal.
## Citation share section
Use this section to understand trust strength behind visibility.
![AI Search Visibility Citation share section with platform bar chart and average share table]()
How to read it:
* Higher citation share means stronger source support.
* Visibility can stay stable while citation support weakens.
* Falling citation share with rising competitors is a priority risk.
## Recent chats section
Use this section to connect aggregate movement to actual runs.
![AI Search Visibility Recent chats section listing recent runs, status, and top results]()
How to use it:
* Find recent runs behind major movement.
* Check recurring top results for source patterns.
* Open key runs to inspect the detail panel shown inside this same view.
## Chat detail panel section
Use this panel when one run needs full evidence-level diagnosis.
![AI Search Visibility chat detail panel with full generated answer and top results list]()
How to use it:
* Read the full answer text first to understand the narrative quality.
* Check `Top results` to see which sources influenced the output.
* Convert findings into concrete prompt, page, or citation actions.
* Recheck the same prompt after updates to confirm improvement.
## Quick weekly checklist
1. Validate top visibility trend direction.
2. Check `Positions` for early ranking-quality losses.
3. Check `Citation share` for trust-support changes.
4. Review `Recent chats` to identify concrete causes.
5. Use `Chat detail` to review evidence for highest-impact runs.
6. Route actions to pages, prompts, and citations.
## How to use filters
* `All prompts` to focused prompt groups for query-level diagnosis.
* `All categories` to isolate funnel-stage behavior.
* Date range to confirm persistent vs temporary movement.
## What to fix first
| Pattern in Visibility data | What it usually means | Recommended action |
| ---------------------------------- | ---------------------------------- | -------------------------------------------------- |
| Visibility down across platforms | Broad relevance loss | Refresh core topic coverage and prompt strategy |
| One platform drops sharply | Platform-specific mismatch | Adjust prompt and content fit for that engine |
| Visibility stable, positions worse | Placement quality is deteriorating | Improve answer relevance and page alignment |
| Citation share down | Trust support weakening | Strengthen citation-ready assets and proof quality |
| Competitor gains on key prompts | Competitive narrative gap | Improve decision content and differentiation |
## Team routine
1. Weekly: diagnose one priority segment.
2. Bi-weekly: compare competitor trend by category.
3. Monthly: report visibility share by platform.
## Keep in mind
* Visibility changes can lead click changes by one cycle.
* Not all visibility has equal business value.
* Segment-level review prevents false conclusions.
## Where to go next
* [Competitors](/data/ai-search/competitors)
* [Prompts](/data/ai-search/prompts)
* [Pages](/data/ai-search/pages)
* [Citations](/data/ai-search/citations)
* [Overview](/data/ai-search/overview)
# Attribution
Source: https://docs.atomicagi.com/data/attribution/overview
See which sources, pages, and countries drive conversions so you can invest in what works
![Attribution overview with conversion trend and the By source attribution tab]()
Use this page to connect traffic activity to conversion outcomes. It helps you decide where to scale, where to fix, and where to stop spending time.
Important: Always compare sessions and conversions
together. Volume without outcomes is not a growth signal.
## Questions this page should answer
1. Which sources are actually contributing to conversions?
2. Which pages create conversion impact, not just visits?
3. Which countries are creating stronger conversion quality?
## Before you analyze
* Keep the same date range you use in Google Search and AI Search reviews.
* Start with a broad view before narrowing to one source, page group, or market.
* Compare against an equivalent previous period.
## What this page gives you
* A top conversion trend with `Total conversions` and `Organic conversions`.
* Three analysis tabs: `By source` for channel-level attribution, `By page` for URL-level attribution, and `By country` for market-level attribution.
* A table under each tab for row-level prioritization.
## How to read the top cards
Start here before going into subtabs:
* `Total conversions`: all conversion outcomes in the selected period.
* `Organic conversions`: conversions tied to organic traffic paths.
How to interpret:
* If total conversions rise but organic conversions fall, non-organic sources may be carrying performance.
* If both rise, your acquisition and conversion quality are aligned.
* If both decline, start with `By source` to find the biggest contributor to the drop.
Key signal: Sessions without conversion growth are usually
a quality problem, not a scale win. Diagnose source and page fit first.
Example: A source can grow from 1,000 to 1,400 sessions
while conversions stay flat. That usually means targeting or landing-page
intent mismatch.
## How these metrics are calculated (simple)
### Total conversions
Total conversions are the selected GA4 conversion events in the chosen date range.
### Organic conversions
Organic conversions are conversion events where source/medium attribution is organic.
### Page value
Page value is a weighted score based on page conversion contribution and session quality signals.
## By source tab
Use this tab to understand which channels and referring sources influence conversions the most.
![Attribution By source tab with source trend and source-level table]()
Focus on:
* `Total sessions` to see source volume.
* `Conversions` to see outcome contribution.
* `Avg. time` as a quality signal.
Take action when:
* A source has high sessions but weak conversions.
* A source has low sessions but very strong conversion rate.
* A source trend changes sharply week over week.
## By page tab
Use this tab to see which landing pages and content URLs drive conversions.
![Attribution By page tab with page trend and page performance table]()
Focus on:
* `Page value` for relative page importance.
* `Total sessions` for traffic contribution.
* `Conversions` for business outcome.
Take action when:
* A high-traffic page under-converts.
* A lower-traffic page over-converts and should be scaled.
* A key commercial page shows stable traffic but conversion decline.
## By country tab
Use this tab to compare conversion performance across markets.
![Attribution By country tab with country trend and country-level table]()
Focus on:
* Which markets produce the largest conversion volume.
* Which markets are improving or declining fastest.
* Whether average engagement differs by country.
Take action when:
* One country shows rising sessions but weak conversion growth.
* A smaller country shows strong conversion efficiency and deserves more focus.
* A priority market loses both sessions and conversions at the same time.
## Quick weekly checklist
1. Check top conversion trend direction first.
2. Review `By source` for biggest winners and losers.
3. Review `By page` for pages to scale or fix this sprint.
4. Review `By country` for market reallocation decisions.
5. Turn insights into clear actions with owner and deadline.
## What to fix first
| Pattern | What it usually means | Recommended action |
| ------------------------------------------- | ---------------------------------------- | -------------------------------------------------------- |
| High sessions, low conversions (source) | Traffic quality mismatch | Rework targeting and landing-page match |
| High page value, falling conversions (page) | Key page is losing conversion efficiency | Refresh offer clarity, CTA placement, and intent match |
| Strong conversion growth in one country | Market-message fit is strong | Replicate campaign and content investment in that market |
| Sessions and conversions both down | Demand or channel distribution issue | Validate campaign mix and content distribution quickly |
## Team routine
1. Weekly: detect source/page/country shifts and assign fixes.
2. Bi-weekly: review conversion efficiency improvements by owner.
3. Monthly: report attribution-based investment changes and outcomes.
## Keep in mind
* Attribution trends are directional, not absolute causality.
* Low-volume segments can be noisy, so validate over multiple periods.
* A conversion increase without engagement quality can be misleading.
## Where to go next
* [Google Search overview](/data/google-search/overview)
* [Google Search landing pages](/data/google-search/landing-pages)
* [AI Search overview](/data/ai-search/overview)
* [Reports](/data/reports/overview)
# Competitor AI Search Visibility
Source: https://docs.atomicagi.com/data/competitors/ai-search-visibility
Compare competitor visibility against citation volume so you can spot who is gaining answer share fastest
![Competitor AI Search visibility page with visibility versus citations chart]()
Use this page to compare competitor visibility in AI answers. It helps you understand who is appearing most often and whether that visibility is backed by stronger citation coverage.
## Questions this page should answer
1. Which competitors own the most AI search visibility?
2. Which competitors combine visibility with strong citations?
3. Which brands are rising fast enough to require a response?
## Before you analyze
* Keep the date range aligned with AI Search Visibility and Citations.
* Focus on the top competitor set first.
* Separate confirmed visibility shifts from speculation about why they happened.
## What this page gives you
* Competitor visibility rows.
* A visibility-versus-citations comparison view.
* Supporting metrics such as position and sentiment when available.
## How to read the visibility chart
Read visibility and citation strength together before deciding what to fix.
* `Visibility`: how often a competitor appears in AI-search answer surfaces.
* `Citations`: how often sources connected to that competitor are cited.
* `Position`: how prominently the competitor appears when it is present.
* `Sentiment`: whether the surrounding answer context is positive, neutral, or negative.
Use this rule:
* High visibility + high citations usually indicates durable authority.
* High visibility + low citations can be a fragile or temporary gain.
* Low visibility + high citations may indicate authority that is not turning into answer share.
## How these metrics are calculated (simple)
```text theme={null}
Visibility = competitor presence across tracked AI-search answer observations
```
```text theme={null}
Citation volume = cited source count associated with competitor visibility
```
Position and sentiment summarize how the competitor appears when present, not why the movement happened.
## How to use this page
### Read visibility and citations together
High visibility with weak citation support can be fragile. High visibility with strong citation support usually indicates a more durable advantage.
### Look for emerging risers
You do not need the market leader to be the biggest immediate risk. A smaller competitor with fast movement often deserves the faster response.
### Use it to choose the next comparison work
When a competitor is gaining here, validate the pages, prompts, and proof patterns behind the move before changing your own content.
## Quick weekly checklist
1. Review the top visibility gainers.
2. Compare each gainer with citation strength.
3. Open related competitor prompts for repeated losses.
4. Check competitor pages for content or proof patterns.
5. Assign one recovery action for the highest-risk competitor movement.
## How to use filters
* Keep the date range aligned with [AI Search visibility](/data/ai-search/visibility).
* Filter by competitor when one brand is moving unusually fast.
* Use prompt/topic filters when visibility changes are tied to one theme.
## What to fix first
| Pattern in competitor visibility | What it usually means | Recommended action |
| --------------------------------------------- | --------------------------------------- | --------------------------------------------------- |
| Competitor visibility rises quickly | They are gaining answer share | Review their prompts, pages, and citations together |
| Competitor has high citations and rank | Their authority signal is strong | Strengthen proof, sources, and comparison content |
| Competitor visibility rises, citations do not | Narrative may be driving the gain | Improve page framing and answer-ready summaries |
| Your brand loses on buying prompts | Commercial pages may be under-supported | Refresh product, comparison, and proof pages |
| Sentiment shifts negative | Answer context may be harming trust | Review source quality and update corrective content |
## Team routine
1. Weekly: check top risers and top losses.
2. Bi-weekly: connect competitor movement to prompt and page fixes.
3. Monthly: report which competitor actions changed visibility or citations.
## Keep in mind
* Visibility alone does not explain the reason for the gain.
* Citation strength often explains whether a competitor advantage is temporary or durable.
* Use the top risers to decide where prompt and content review should start.
## Where to go next
* [Competitors overview](/data/competitors/overview)
* [Competitor prompts](/data/competitors/prompts)
* [AI Search visibility](/data/ai-search/visibility)
* [Citations](/data/ai-search/citations)
# Competitors
Source: https://docs.atomicagi.com/data/competitors/overview
Track accepted competitors, crawl status, and suggested additions so your competitive coverage stays current
![Competitors page with tracked competitor domains and crawl status]()
Use this page to maintain the competitor set behind the rest of the competitor views. This is where you decide which domains belong in your ongoing competitive analysis.
## Questions this page should answer
1. Which competitor domains are currently tracked?
2. Which suggested competitors should we accept or dismiss?
3. Which competitor crawls are ready, pending, or failed?
## Before you manage competitors
* Keep the list focused on direct search competitors, not every adjacent brand.
* Add domains in batches when you are expanding into a new segment.
* Review crawl status before expecting data on the other competitor pages.
## What this page gives you
* The accepted competitor list for the project.
* Suggested competitors you can accept or dismiss.
* Crawl status and crawl recency for each competitor.
* A form to add domains manually.
## How to read the competitor list
Read this page as the source of truth for every downstream competitor report.
* `Domain`: the competitor Atomic will compare against your project.
* `Status`: whether the competitor crawl is ready, pending, or failed.
* `Last crawled`: how fresh the competitor page data is.
* `Suggested competitors`: candidate domains that still need review.
Use this rule:
* If a competitor is not accepted here, do not expect it in competitor page or prompt analysis.
* If a competitor crawl is pending, wait before diagnosing missing downstream data.
* If a competitor is no longer strategically relevant, remove it before it pollutes comparisons.
## How these statuses work (simple)
```text theme={null}
Accepted competitor = domain selected for ongoing project comparison
```
```text theme={null}
Crawl status = latest collection state for that competitor domain
```
Suggested competitors are discovery candidates. They become part of analysis only after someone accepts them.
## How to use this page
### Keep the list narrow and intentional
Too many competitors create noisy page and prompt comparisons. Track the domains that compete for the same audience, SERP space, or AI answer share.
### Review suggestions before accepting
Suggested competitors save time, but they still need human judgment. Accept the domains that truly overlap with your offer and dismiss the rest.
### Watch crawl status
Pending or failed crawls explain missing downstream competitor data. Fix the tracked list first before assuming the analysis pages are wrong.
## Quick weekly checklist
1. Review new suggested competitors.
2. Accept only domains that compete for the same audience, SERP, or AI-answer space.
3. Remove stale competitors that no longer match the strategy.
4. Check crawl status before opening page-level or prompt-level competitor reports.
5. Re-run downstream analysis after large competitor list changes.
## How to use filters
* Filter by status when you need to find pending or failed crawls quickly.
* Search by domain when the competitor list is large.
* Review suggestions in batches so the accepted list stays intentional.
## What to fix first
| Pattern on Competitors page | What it usually means | Recommended action |
| ---------------------------------- | ------------------------------------- | ------------------------------------------------------ |
| Many suggestions are unreviewed | Competitor set is drifting | Accept or dismiss suggestions before deeper analysis |
| Important competitor is missing | Downstream reports will under-compare | Add the domain manually and wait for crawl completion |
| Crawl status is failed or stale | Page/prompt data may be incomplete | Re-check domain validity and retry crawl if available |
| Too many accepted competitors | Analysis will become noisy | Keep only direct strategic or search competitors |
| Competitors are from mixed markets | Benchmarks will be misleading | Split analysis by market or keep only relevant domains |
## Team routine
1. Weekly: review suggestions and crawl status.
2. Bi-weekly: prune accepted competitors that no longer matter.
3. Monthly: align the competitor set with campaign, market, and AI-search priorities.
## Keep in mind
* This page controls the competitor set used by the other competitor views.
* New competitors may need time before page-level analysis becomes useful.
* A clean competitor list improves both prompt and content-gap decisions.
## Where to go next
* [Competitor pages](/data/competitors/pages)
* [Competitor AI search visibility](/data/competitors/ai-search-visibility)
* [Competitor prompts](/data/competitors/prompts)
* [AI Search competitors](/data/ai-search/competitors)
# Competitor Pages
Source: https://docs.atomicagi.com/data/competitors/pages
Review competitor URLs and coverage gaps so you can find missing content areas and overlapping coverage faster
![Competitor Pages view with URL-level coverage rows across tracked competitors]()
Use this page to inspect competitor URLs across the current tracked set. It helps you see where competitors cover topics you do not, and where your own pages already overlap.
## Questions this page should answer
1. Which competitor pages are uncovered by our site?
2. Which competitor pages already have a comparable page on our site?
3. Which page-level gaps are large enough to prioritize?
## Before you analyze
* Confirm the competitor list is current.
* Check crawl status on the main Competitors page first.
* Use this page after you know which competitors matter.
## What this page gives you
* URL-level competitor coverage rows.
* Coverage and change signals for each page.
* The matched own URL when a comparable page already exists.
## How to read the page table
Review the table by theme, not by one isolated URL.
* `Competitor URL`: the page Atomic found on a tracked competitor.
* `Matched own URL`: your closest comparable page, when one exists.
* `Coverage`: whether your site already covers the same intent.
* `Change signal`: whether competitor coverage is new or moving.
Use this rule:
* A single unmatched page is a research prompt.
* Repeated unmatched pages in one topic are a content gap.
* Matched pages with weak performance usually need a refresh, not a new page.
## How these metrics are calculated (simple)
```text theme={null}
Coverage gap = competitor page intent without a strong matching page on your site
```
```text theme={null}
Matched own URL = closest project URL Atomic associates with the competitor page intent
```
The table is a prioritization aid. Human review still decides whether the gap needs a new page, a section update, or no action.
## How to use this page
### Start with unmatched coverage
Pages with no comparable page on your site usually represent the clearest content gap. Review those before spending time on already-covered areas.
### Validate the reason, not just the row
Some gaps need a new page. Others only need a section, a stronger comparison block, or better internal linking.
### Watch recent changes
If competitor coverage changed recently, the page may reveal a new cluster or a fresh push from a direct rival.
## Quick weekly checklist
1. Filter to your highest-priority competitor or topic.
2. Review unmatched pages first.
3. Group repeated gaps into themes.
4. Check whether an existing page can be refreshed before creating a new page.
5. Turn the strongest gap into one content or optimization task.
## How to use filters
* Use competitor filters to inspect one rival at a time.
* Use topic or search filters to avoid mixing unrelated gaps.
* Use status/coverage filters to separate unmatched pages from already-covered URLs.
## What to fix first
| Pattern in Competitor Pages | What it usually means | Recommended action |
| --------------------------------- | --------------------------------- | --------------------------------------------------- |
| Many unmatched pages in one theme | Real content coverage gap | Create or refresh one strategic page cluster |
| One unmatched page only | Possible edge case | Validate search demand before creating new content |
| Matched page exists but is weak | Existing page needs improvement | Refresh title, structure, proof, and internal links |
| Competitor changed many URLs | Rival is investing in a topic | Review prompt visibility and content depth together |
| Match looks wrong | Intent mapping needs human review | Inspect both pages before assigning work |
## Team routine
1. Weekly: review new or high-impact gaps.
2. Bi-weekly: convert repeated gaps into content refresh tasks.
3. Monthly: compare gap work with prompt and AI-search visibility movement.
## Keep in mind
* URL-level gaps do not automatically justify a new page.
* Repeated uncovered patterns are stronger signals than one-off pages.
* Use this page together with Prompts and AI search visibility for prioritization.
## Where to go next
* [Competitors overview](/data/competitors/overview)
* [Competitor prompts](/data/competitors/prompts)
* [Competitor AI search visibility](/data/competitors/ai-search-visibility)
* [Opportunities](/data/opportunities/overview)
# Competitor Prompts
Source: https://docs.atomicagi.com/data/competitors/prompts
Compare tracked prompts across competitors so you can see where your brand loses share on the questions that matter most
![Competitor Prompts page comparing tracked prompts across competitors]()
Use this page to compare tracked prompts across your competitor set. It helps you identify where competitor narratives are stronger on the exact prompts your team cares about.
## Questions this page should answer
1. Which competitors are strongest on our tracked prompts?
2. Which prompt groups are we losing most often?
3. Which prompt losses are worth fixing first?
## Before you analyze
* Make sure your tracked prompts are current.
* Filter to one topic or prompt family at a time when diagnosing a loss.
* Pair prompt movement with citations before deciding the fix.
## What this page gives you
* Prompt-level competitor comparisons.
* A way to isolate which brands lead on each tracked question.
* Clear input for content, proof, and authority decisions.
## How to read the prompt comparison table
Start with prompt groups tied to business outcomes.
* `Prompt`: the tracked question or answer scenario.
* `Competitor`: the brand appearing or performing strongly for that prompt.
* `Visibility/rank`: how often and how prominently the competitor appears.
* `Citation/support`: the source evidence behind the answer when available.
Use this rule:
* A loss on one prompt is a signal to inspect.
* Repeated losses across a prompt family are a strategy issue.
* A competitor win with strong citations usually needs proof and authority work, not just copy edits.
## How these metrics are calculated (simple)
```text theme={null}
Prompt win = competitor appears more strongly than your brand on a tracked prompt
```
```text theme={null}
Prompt cluster risk = repeated prompt wins for the same competitor or topic
```
Use prompt comparisons with citations and visibility before deciding the fix.
## How to use this page
### Start with business-critical prompts
Do not spread effort across every prompt. Focus first on the prompts tied to buying intent, comparisons, or product category definition.
### Separate message loss from authority loss
If a competitor wins prompts but not citations, your framing may be weak. If they win both, you likely need stronger proof and stronger pages.
### Turn prompt losses into focused work
One prompt cluster should lead to one concrete next action: improve a page, add proof, create a comparison asset, or strengthen external citations.
## Quick weekly checklist
1. Filter to high-value prompt families.
2. Identify repeated competitor wins.
3. Check whether the competitor also has stronger citations.
4. Decide whether the fix is page copy, proof, comparison content, or authority work.
5. Create one task per prompt cluster instead of one task per prompt.
## How to use filters
* Filter by competitor when diagnosing a known rival.
* Filter by prompt family or topic to avoid mixing unrelated intent.
* Keep date range aligned with AI Search Visibility and Citations.
## What to fix first
| Pattern in Competitor Prompts | What it usually means | Recommended action |
| --------------------------------- | ----------------------------------------- | ------------------------------------------------- |
| Same competitor wins many prompts | They own the narrative for that theme | Refresh core page messaging and proof |
| Competitor wins only one prompt | Could be isolated noise | Watch before creating large work |
| Competitor wins and has citations | Authority gap supports their answer share | Build proof, citations, and source-backed content |
| Your page exists but still loses | Page may not answer the prompt clearly | Add direct answer sections and comparison framing |
| Prompt list is stale | Analysis is no longer aligned to strategy | Refresh tracked prompts before assigning work |
## Team routine
1. Weekly: review top prompt losses on commercial themes.
2. Bi-weekly: update content tasks based on repeated prompt clusters.
3. Monthly: refresh tracked prompts and prune low-value questions.
## Keep in mind
* Prompt-level losses are easier to act on than broad visibility shifts.
* Competitor wins on a single prompt matter less than repeated wins across a theme.
* This page is most useful when prompts are curated intentionally.
## Where to go next
* [Prompts](/data/ai-search/prompts)
* [Competitor AI search visibility](/data/competitors/ai-search-visibility)
* [Competitors overview](/data/competitors/overview)
* [Mentions](/data/ai-search/mentions)
# Devices
Source: https://docs.atomicagi.com/data/google-search/devices
Compare mobile, desktop, and device-level performance
![Google Search devices tab with device trend and breakdown table]()
Use this page to compare performance by device and decide where to focus first. It helps you answer three practical questions:
1. Is growth or decline concentrated on one device type?
2. Are ranking and visibility changes different on mobile vs desktop?
3. Which device should drive your next optimization sprint?
Important: Device movement often points to template or UX
issues. Pair device data with landing-page and recent-release context.
## Before you analyze
* Keep the same date window when comparing devices.
* Check absolute volume first, then percentage movement.
* Check device shifts against recent page or template releases.
## What this page gives you
The Devices page combines:
* Device trend lines for `Desktop`, `Mobile`, and `Tablet`.
* Summary metrics per device to compare contribution and momentum.
* A compact breakdown table with `Clicks`, `Impressions`, and `Conversions`.
* Shared filters for keyword, page, device, and country segmentation.
## How to read device performance correctly
* `Clicks` tells you current traffic contribution by device.
* `Impressions` shows demand and visibility by device.
* `Conversions` helps avoid optimizing only for traffic volume.
A device with lower traffic but stronger conversions may deserve higher priority than a high-volume but weak segment.
## How these metrics are calculated (simple)
```text theme={null}
Device row = Google Search metrics grouped by device type
```
```text theme={null}
CTR = clicks / impressions
```
Conversions come from connected analytics data where attribution is available for the selected period.
## Quick weekly checklist
1. Read the overall trend.
Check whether all devices move together or one diverges.
2. Compare contribution levels.
Identify the dominant device and the fastest mover.
3. Evaluate conversion quality.
Avoid over-investing in segments with weak downstream impact.
4. Link to release timeline.
Match device shifts to recent page, template, or UX changes.
5. Convert into focused tasks.
Assign mobile, desktop, or tablet-specific fixes and owners.
Use the device breakdown table for quick decision-making in weekly operations.
![Google Search devices breakdown table with clicks impressions and conversions per device]()
## How to use filters
Use filters to isolate root cause before assigning work.
![Google Search devices filters panel]()
* `Device = Mobile`
Isolate mobile behavior and validate UX-driven ideas.
* `Page contains [/blog/ or /product/]`
Compare template-level device performance differences.
* `Country = [market]`
Check whether device behavior shifts by region.
* `Keyword contains [cluster term]`
Detect topic-specific differences in intent and CTR by device.
## What to fix first
| Pattern in device data | What it usually means | Recommended action |
| --------------------------------------------- | -------------------------------- | -------------------------------------------------------- |
| Mobile down, desktop stable | Mobile UX or rendering issue | Audit mobile templates, performance, and search snippets |
| Impressions up, clicks flat on one device | CTR is weak | Improve titles/meta and intent alignment |
| Clicks stable, conversions down on one device | On-page friction after the click | Review layout, CTA placement, and forms |
| Tablet volatile with low baseline | Low sample-size noise | Monitor longer before high-effort changes |
| Desktop up, mobile down after redesign | Responsive layout regression | Run targeted QA and fix breakpoint-specific issues |
## Team routine
1. Weekly: flag largest device divergence and assign corrective tasks.
2. Bi-weekly: validate if fixes improved target device metrics.
3. Monthly: report device split trend and conversion quality changes.
4. Quarterly: review device-specific template strategy and performance debt.
## Keep in mind
* Tablet data can be sparse; treat large percentages carefully.
* Device type alone does not explain intent; pair with page and keyword segments.
* Short windows can overstate noise, especially on smaller traffic sources.
## Where to go next
* [Google Search overview](/data/google-search/overview): top-level trend context
* [Landing pages](/data/google-search/landing-pages): page-level impact by device behavior
* [Keywords](/data/google-search/keywords): query intent differences behind device shifts
* [Geography](/data/google-search/geography): device performance differences across markets
# Geography
Source: https://docs.atomicagi.com/data/google-search/geography
Compare performance by market and region
![Google Search geography tab with world map and country table]()
Use this page to make country-level SEO decisions with clear data. It helps you answer three key questions:
1. Which markets are driving growth or creating drag?
2. Where are impressions high but clicks weak?
3. Which countries deserve localized content or technical focus next?
Important: Do not prioritize countries by traffic alone.
Use conversion quality and ranking direction before assigning localization
work.
## Before you analyze
* Use comparable date ranges before judging market movement.
* Start globally, then isolate one region at a time.
* Review both volume (`Clicks`, `Impressions`) and quality (`Position`, `Conversions`).
## What this page gives you
The Geography page combines:
* A world map heat layer to show click concentration by country.
* A country table with `Clicks`, `Impressions`, `Position`, and `Conversions`.
* View toggles in the map area for different perspectives.
* Shared filters (`Keyword`, `Page`, `Device`, `Country`) for focused regional analysis.
## How to read geography metrics correctly
* `Clicks`: current traffic contribution by market.
* `Impressions`: search demand and visibility footprint in that country.
* `Position`: how strong your ranking is in local search.
* `Conversions`: helps you avoid over-prioritizing low-intent markets.
A market with lower clicks but strong conversions may deserve more investment than a high-volume market with weak outcomes.
## How these metrics are calculated (simple)
```text theme={null}
Country row = Google Search metrics grouped by searcher country
```
```text theme={null}
CTR = clicks / impressions
```
Conversions come from connected analytics data where attribution is available for the selected period.
## Quick weekly checklist
1. Start with map concentration.
Check if performance depends on a small number of countries.
2. Scan top-country table rows.
Find biggest movers in clicks and conversions first.
3. Check ranking efficiency by market.
Prioritize countries where position is improving but clicks lag.
4. Compare mature vs emerging markets.
Separate stabilization tasks from expansion experiments.
5. Convert findings into localized actions.
Assign language updates, regional content, and market-specific SEO tasks.
Use deeper rows to find mid-tier markets that can become your next growth layer.
![Google Search geography table showing deeper country rows and performance deltas]()
## How to use filters
Use filters to test one country-specific idea before planning execution.
![Google Search geography filters panel]()
* `Country = [target market]`
Isolate one market to check local trend quality.
* `Page contains [/blog/ or /product/]`
Compare template performance by country intent.
* `Keyword contains [cluster term]`
Verify if a topic performs unevenly across markets.
* `Device = Mobile`
Detect regional mobile UX/ranking gaps.
## What to fix first
| Pattern in geography data | What it usually means | Recommended action |
| -------------------------------------------- | -------------------------------------------- | ----------------------------------------------------------- |
| High impressions, low clicks in one country | Local messaging does not match search intent | Localize titles/meta and improve local messaging |
| Clicks down, position worse in a key market | More competition or weaker market fit | Refresh localized pages and strengthen local internal links |
| Position improving, conversions flat | More visibility but weaker traffic quality | Adjust content intent and conversion path |
| Small market, strong conversion rate | High-quality niche opportunity | Expand targeted content and supporting pages |
| Large market, volatile week-to-week movement | Seasonality or campaign noise | Validate with longer windows before reallocating resources |
## Team routine
1. Weekly: identify largest country-level gains/losses and assign owners.
2. Bi-weekly: review conversion quality shifts in top markets.
3. Monthly: publish regional scorecard with localized action plan.
4. Quarterly: decide where to expand or reduce localization investment.
## Keep in mind
* Country-level averages can hide city-level differences.
* Position is an average and not a single tracked keyword rank.
* New market growth can be noisy; confirm persistence over multiple periods.
## Where to go next
* [Google Search overview](/data/google-search/overview): macro trend context
* [Keywords](/data/google-search/keywords): query-level differences by market
* [Landing pages](/data/google-search/landing-pages): URL-level market outcomes
* [Devices](/data/google-search/devices): mobile/desktop split by geography
# Keywords
Source: https://docs.atomicagi.com/data/google-search/keywords
Find which queries are driving growth and which need optimization
![Google Search keywords tab with trend and table]()
Use this page to see which keywords are helping you grow and which ones need work. It helps you answer three practical questions:
1. Which keyword groups are going up or down?
2. Are changes caused by ranking movement, search demand, or CTR?
3. Which queries should be prioritized in this sprint?
Important: Separate branded, non-branded, and question
intent before assigning keyword work. Mixed intent creates misleading
priorities.
## Before you analyze
* Set a clear date range (weekly for execution, monthly for strategy).
* Reset filters unless you are intentionally reviewing one segment.
* Compare similar periods so seasonality does not mislead you.
## What this page gives you
The Keywords page combines:
* Keyword distribution cards (`Top 3`, `4-10`, `11-50`) to show ranking health.
* A trend chart for clicks, impressions, position, and CTR.
* A keyword table for row-level prioritization by query.
* Segment tabs (`Totals`, `Individual keywords`, `Branded keywords`, `Questions`) for different analysis modes.
## How to read the top cards
* `Total organic keywords`: total number of keywords you rank for.
* `Top 3 positions`: strongest visibility and highest click potential.
* `4-10 positions`: immediate page-one opportunities.
* `11-50 positions`: keywords that need more content depth and linking support.
If `Top 3` drops while `11-50` grows, you are usually losing rankings instead of gaining stronger visibility.
## How these metrics are calculated (simple)
```text theme={null}
Keyword bucket = count of tracked queries in a ranking position range
```
```text theme={null}
CTR = clicks / impressions
```
Average position is calculated across the selected keywords, pages, date range, and filters.
## Individual keywords tab
Use this tab when you want to prioritize exact queries, one by one.
![Google Search Individual keywords tab with query-level trends and table]()
Focus on high-impression queries first, then decide the action:
* Good position, weak CTR: improve title and meta copy.
* Weak position, stable impressions: improve page depth and internal links.
* Rising impressions, flat clicks: adjust content to match search intent better.
## Branded keywords tab
Use this tab to separate brand demand from non-brand SEO performance.
![Google Search Branded keywords tab with brand query performance]()
This is useful for clean reporting:
* If branded is stable but totals drop, the problem is usually non-brand visibility.
* If branded drops hard, check campaigns, PR activity, and brand SERP competition.
* Keep branded and non-branded actions separate so priorities stay clear.
## Questions tab
Use this tab to find question-based queries you can win with clearer answers.
![Google Search Questions tab with question-intent keyword performance]()
This tab helps with content updates:
* Find question keywords with impressions but low clicks.
* Add short, direct answers near the top of the matching page.
* Expand FAQ sections and structure headings around real question phrasing.
## Quick weekly checklist
1. Check trend direction first.
Compare clicks and impressions before reading CTR.
2. Review position movement.
Rising average position value with stable impressions usually means ranking loss.
3. Scan the table for high-impact rows.
Prioritize high-impression keywords with weak CTR or declining clicks.
4. Split by tab.
Use `Branded` and `Questions` to avoid mixing intent and demand patterns.
5. Create actions by keyword group.
Map each target query to a page update, internal link change, or new supporting content.
The lower part of the table is where new and long-tail opportunities often appear.
![Google Search keywords table with deeper rows and row-level metrics]()
## How to use filters
Use filters to test one idea at a time before creating tasks.
![Google Search keywords filters panel]()
* `Keyword contains [cluster term]`
Check one topic area without noise from unrelated keywords.
* `Page contains [/blog/ or /product/]`
Separate informational vs commercial intent performance.
* `Device = Mobile`
Diagnose mobile-specific CTR and ranking differences.
* `Country = [market]`
Check whether changes are global or localized.
## What to fix first
| Pattern in table | What it usually means | Recommended action |
| --------------------------------------------- | --------------------------------------------- | --------------------------------------------------------- |
| High impressions, low CTR | Your result is visible but not persuasive | Rewrite title and meta description to better match intent |
| Clicks down, impressions flat, position worse | Rankings are dropping | Refresh page content and strengthen internal links |
| Position improving, CTR flat | Better rankings but weak click appeal | Test clearer value in title and meta |
| New impressions, low clicks on question terms | Early visibility, but page answers are weak | Expand FAQ sections and add concise answers |
| Branded stable, non-branded declining | Non-brand topics are getting more competitive | Expand cluster content and build topic authority |
## Team routine
1. Weekly: identify top gains/losses and assign actions.
2. Bi-weekly: compare branded vs non-branded trends.
3. Monthly: report cluster movement and completed updates.
4. Quarterly: review `11-50` keywords for new content opportunities.
## Keep in mind
* Google Search Console is delayed, so avoid same-day conclusions.
* Average position combines many URLs and keywords.
* A CTR drop is not always a copy issue; intent can change too.
## Where to go next
* [Google Search overview](/data/google-search/overview): diagnose top-level movement
* [Landing pages](/data/google-search/landing-pages): turn query findings into URL-level actions
* [Geography](/data/google-search/geography): validate country-specific demand and rank shifts
* [Devices](/data/google-search/devices): isolate mobile vs desktop behavior
# Landing pages
Source: https://docs.atomicagi.com/data/google-search/landing-pages
Understand which pages are winning and which pages need attention
![Google Search landing pages tab with trend and page table]()
Use this page to see which URLs are helping growth and which need updates first. It helps you answer three key questions:
1. Which pages are driving or losing organic traffic right now?
2. Are page changes caused by ranking movement, demand shifts, or CTR issues?
3. Which pages should be updated first this sprint?
Important: Diagnose page movement before rewriting. Demand,
rankings, snippets, and technical issues can produce the same traffic
pattern.
## Before you analyze
* Set a date range that matches your reporting schedule.
* Start with no filters to see the global pattern first.
* Compare against an equivalent prior period (week-over-week or month-over-month).
## What this page gives you
The Landing pages view combines:
* A trend chart for clicks, impressions, position, and CTR.
* A page-level table with performance deltas by URL.
* Two modes:
* `Total pages` for broad portfolio movement.
* `Pages` for detailed row-level prioritization.
* The same filter stack used across Google Search views.
## How to read the table correctly
* `Clicks`: pages generating organic visits now.
* `Impressions`: pages with visibility potential, even when clicks are low.
* `Position`: average ranking level for searches that lead to that page.
Prioritize pages where impressions stay high but clicks drop. These are usually easier wins.
## How these metrics are calculated (simple)
```text theme={null}
Page row = Google Search Console metrics grouped by landing page URL
```
```text theme={null}
CTR = clicks / impressions
```
Position is the average ranking position for queries that led to that page during the selected period.
## Quick weekly checklist
1. Start with trend direction.
Check if the change is site-wide or limited to a few pages.
2. Sort by clicks change.
Identify the biggest traffic movers first.
3. Check position on losing pages.
Separate ranking decay from demand changes.
4. Group by template or content type.
Example: `/blog/`, `/industry-updates/`, `/product/`.
5. Turn findings into action tickets.
Assign owner, URL, change type, and expected impact window.
For deeper review, scan lower rows where mid-volume opportunities usually sit.
![Google Search landing pages deeper table rows for URL-level prioritization]()
## How to use filters
Use filters to test one idea at a time before creating tasks.
![Google Search landing pages filters panel]()
* `Page contains [/blog/ or /industry-updates/]`
Compare performance by content type and template.
* `Keyword contains [topic term]`
Check whether a specific topic is driving movement.
* `Device = Mobile`
Identify mobile-first page losses and UX-sensitive URLs.
* `Country = [market]`
Check if winners and losers are regional rather than global.
## What to fix first
| Pattern in page table | What it usually means | Recommended action |
| ----------------------------------------- | ----------------------------------------- | --------------------------------------------------------- |
| High impressions, low or declining clicks | People see the page but do not click | Rewrite title/meta and align messaging with search intent |
| Clicks down and position worsened | Rankings are slipping | Refresh page content and improve internal links |
| Impressions up, clicks flat | Visibility is up, but click appeal is low | Improve search snippet and strengthen the page intro |
| New page with fast impression growth | Google is testing this page more often | Expand depth and add related internal links |
| Stable position but traffic down | Demand or keyword mix changed | Recheck page intent and topic coverage |
## Team routine
1. Weekly: top losers/winners and immediate fixes.
2. Bi-weekly: cluster-level review by URL groups.
3. Monthly: report recovered pages, still-declining pages, and next experiments.
4. Quarterly: identify pages to consolidate, merge, or fully refresh.
## Keep in mind
* Search Console data is delayed, so avoid same-day conclusions.
* Average position is an overall average, not one fixed rank.
* One page can rank for different intents, so check keyword data too.
## Where to go next
* [Google Search overview](/data/google-search/overview): macro trend diagnosis
* [Keywords](/data/google-search/keywords): query-level explanation behind page movement
* [Geography](/data/google-search/geography): country-level effects on page traffic
* [Devices](/data/google-search/devices): mobile vs desktop page performance
# Google Search Overview
Source: https://docs.atomicagi.com/data/google-search/overview
Understand organic performance fast, diagnose change, and decide the next SEO actions
![Google Search overview trend chart]()
Use this page as your starting point for Google Search performance. It helps you answer three core questions:
1. Is organic visibility moving in the right direction?
2. What is driving the change: demand, rankings, or click-through rate?
3. What should we do next this week?
Important: Read clicks, impressions, CTR, and position
together. One metric by itself rarely explains why organic performance
changed.
## Before you analyze
* Make sure Google Search Console is connected to the project.
* Pick a date range that matches your reporting schedule.
* Compare equivalent periods before concluding performance changed structurally.
## What this page gives you
The overview combines:
* A top-level trend chart for key organic metrics.
* Fast previews of the most important breakdown tables.
* Shared filters (`Keyword`, `Page`, `Device`, `Country`) so you can narrow the data before drilling down.
The scrollable lower section is where you confirm what is driving movement across pages, countries, and devices.
![Google Search overview lower section with pages countries and devices]()
## How to read each metric correctly
| Metric | What it tells you | Common misread |
| ----------- | ------------------------------------------------------------------- | ----------------------------------------------------------------------------------- |
| Clicks | Organic visits from Google Search | Treating short-term swings as permanent change without checking other metrics first |
| Impressions | How often your pages were shown in Google results | Assuming impressions always mean qualified traffic |
| CTR | How often people click after seeing your result | Reading CTR without checking position and query mix |
| Position | Average ranking across included queries and pages (lower is better) | Treating average position as a single keyword rank |
## How these metrics are calculated (simple)
```text theme={null}
CTR = clicks / impressions
```
```text theme={null}
Position = average ranking position across included Google Search queries and pages
```
Clicks and impressions come from Google Search Console for the selected date range and filters.
## Quick weekly checklist
1. Start with trend direction.
Confirm whether clicks and impressions are moving up, flat, or down.
2. Check CTR and position movement.
Separate ranking issues (`Position`) from snippet/intent issues (`CTR`).
3. Scan preview tables.
Look for concentration risk: a small set of pages, countries, or devices driving most movement.
4. Segment with filters.
Narrow to one country or device when movement is not uniform.
5. Assign next actions.
Route issues to content, technical SEO, internal linking, or market-specific teams.
## How to use filters
![Google Search overview filters panel]()
* `Keyword contains [topic cluster]`
Check if a topic initiative is improving visibility.
* `Page contains [/blog/ or /product/]`
Compare template performance and find weak URL groups.
* `Device = Mobile`
Check for mobile-specific issues before prioritizing fixes.
* `Country = [target market]`
Use this for international analysis and market-specific planning.
## What to fix first
| Signal in overview | What it usually means | Go next |
| ----------------------------------------- | --------------------------------------------------------- | ------------------------------------------------------------------------------------------------------- |
| Impressions up, CTR down | Pages are shown more, but snippets are not getting clicks | [Keywords](/data/google-search/keywords), then [Landing pages](/data/google-search/landing-pages) |
| Impressions down, position down | Ranking loss, technical issues, or content getting stale | [Landing pages](/data/google-search/landing-pages), then [Technical overview](/data/technical/overview) |
| Clicks down, impressions flat | Click-through rate dropped on important queries | [Keywords](/data/google-search/keywords) |
| Mobile down, desktop stable | Likely mobile UX or template issue | [Devices](/data/google-search/devices), then [Landing pages](/data/google-search/landing-pages) |
| One country drops while others are stable | Regional competition or localization gap | [Geography](/data/google-search/geography) |
## Team routine
Use this structure in weekly client updates:
1. `Performance summary`: "Organic clicks are \[up/down] by X% for \[period]."
2. `Driver summary`: "Main change comes from \[queries/pages/country/device]."
3. `Action plan`: "This week we will \[titles/meta refresh, content update, internal links, technical fix]."
4. `Impact window`: "We expect movement in \[2-6] weeks depending on crawl speed and index updates."
## Keep in mind
* Google Search Console data is delayed by 1-3 days.
* Position is an average across many queries and URLs.
* Seasonality and campaign timing can change demand; validate against comparable periods.
## Where to go next
* [Keywords](/data/google-search/keywords): query-level opportunities and losses
* [Landing pages](/data/google-search/landing-pages): URL-level winners and declines
* [Geography](/data/google-search/geography): country-level performance shifts
* [Devices](/data/google-search/devices): device split and mobile/desktop issues
* [Referrals](/data/google-search/referrals): source-level context beyond rankings
# Referrals
Source: https://docs.atomicagi.com/data/google-search/referrals
Understand referral patterns tied to organic discovery
![Google Search referrals tab with domain trend and source table]()
Use this page to see which external sites are sending traffic to your content. It helps you answer three practical questions:
1. Which referral domains are growing and sending useful traffic?
2. Which sources are declining and need replacement or recovery?
3. How should SEO, PR, and distribution priorities change based on source movement?
Important: Referral spikes are not automatically durable
channels. Check repeat contribution and source quality before scaling
effort.
## Before you analyze
* Use a stable comparison window so short-term spikes do not mislead you.
* Start with top sources, then scan mid-tier sources for emerging channels.
* Review both session volume and percentage change before setting priorities.
## What this page gives you
The Referrals page combines:
* A `Referring domains` trend card to show how your source mix changes over time.
* A sources table with `Total sessions` and directional deltas.
* Source labels in rows (`new`, `lost`) so you can track churn.
## How to read this page correctly
* `Referring domains` up with flat sessions can mean broader but shallower discovery.
* Sessions up from a small set of domains can signal concentration risk.
* High growth percentages on tiny baselines should not outweigh high-volume stable sources.
## How these metrics are calculated (simple)
```text theme={null}
Referring domains = unique external domains sending sessions in the selected period
```
```text theme={null}
Total sessions = sessions attributed to each source for the selected date range
```
`New` and `lost` labels compare source presence against the previous comparison period.
## Quick weekly checklist
1. Start with domain trend.
Confirm whether source footprint is expanding or shrinking.
2. Review top-volume sources.
Validate if core referrers are stable enough to rely on.
3. Scan fast movers.
Separate meaningful emerging channels from one-off spikes.
4. Inspect `new` and `lost` labels.
Catch partner drop-offs, platform shifts, or campaign changes.
5. Turn findings into an action list.
Assign source-level actions to SEO, content distribution, and partnerships.
Use this middle table zone for weekly source prioritization.
![Google Search referrals table with top sources and session change]()
## What to fix first
| Pattern in source table | What it usually means | Recommended action |
| -------------------------------------------- | ----------------------------------- | ------------------------------------------------------ |
| High sessions, positive growth | Reliable source channel | Protect and scale with repeat placements |
| High sessions, negative growth | Channel fatigue or lower visibility | Refresh distribution hooks and update linking assets |
| Low sessions, very high growth | Early source opportunity | Run small tests before scaling effort |
| `new` sources with repeat sessions | New channels are starting to work | Build repeatable collaboration or syndication workflow |
| `lost` sources with prior meaningful traffic | Broken loop or campaign drop-off | Investigate and create a recovery or replacement plan |
## Track new and lost sources
This deeper table view helps you spot source churn that top-level numbers can hide.
![Google Search referrals movement view with new and lost source labels]()
* Rising `new` sources with low volume: validate quality before scaling.
* Repeated `lost` labels on formerly strong domains: trigger recovery workflow.
* Mixed trend (many new + many lost): treat as channel instability, not pure growth.
## Team routine
1. Weekly: identify top 5 gaining and top 5 declining sources.
2. Bi-weekly: map source changes to recent content or campaign launches.
3. Monthly: report referral mix, dependency risk, and next channel experiments.
4. Quarterly: rebalance channel strategy based on sustained referral contribution.
## Keep in mind
* Referral traffic can spike from short-lived mentions; confirm it lasts.
* Percentage growth without baseline context can mislead prioritization.
* Not every referral source matches your target intent; align quality with conversions.
## Where to go next
* [Google Search overview](/data/google-search/overview): top-level organic movement context
* [Landing pages](/data/google-search/landing-pages): which URLs are benefiting from referral traffic
* [Keywords](/data/google-search/keywords): queries behind the pages that sources amplify
* [Geography](/data/google-search/geography): market-level differences in referral impact
# Opportunities groups
Source: https://docs.atomicagi.com/data/opportunities/groups
Group opportunities by theme so teams can prioritize the right workstream first
![Opportunity groups overview with summary cards and grouped issue tables]()
Use this page to decide which issue themes your team should fix first, then drill into one group and launch AI help.
Important: A high Action items count does not always mean
highest priority. Use Impact and business relevance first.
## Questions this page should answer
1. Which opportunity groups create the biggest upside right now?
2. Which group should we open first for detailed URL-level work?
3. When should we use `Fix with AI` to speed up execution?
## Before you analyze
* Keep the same project and date context you use in other Opportunities pages.
* Start with `All impact levels`, then narrow once priorities are clear.
* Pick one high-impact group first before assigning tasks.
## What this page gives you
* Top workflow cards: `All items`, `Unresolved`, `In review`, `Resolved`.
* Grouped issue tables (for example `Content visibility` and `Technical`).
* For each group: `Issue`, `Effort`, `Impact`, `Action items`, and `Last checked`.
* One-click drill-down into affected URLs and fix guidance.
## How to read the top cards
* `All items`: full grouped opportunity count in current scope.
* `Unresolved`: items still needing execution.
* `In review`: items currently being validated.
* `Resolved`: items marked complete.
Use these cards to track workflow state, then prioritize inside the grouped issue tables.
## How these metrics are calculated (simple)
### Impact
Impact is the estimated SEO/business upside if that issue group is resolved.
### Effort
Effort is the relative implementation complexity and time needed to resolve the group.
### Action items
Action items are the count of affected URL-level issues inside that group.
## Step 1: Start from the opportunity groups overview
In the overview, focus on high-impact groups with meaningful `Action items` counts. This gives you the best first target to open.
* `High impact` + lower effort usually means faster wins.
* Larger `Action items` counts usually indicate a broader issue pattern.
* `Last checked` helps you avoid acting on stale groups first.
## Step 2: Open one group to see affected pages and issue details
When you click a group row, a modal opens with URL-level items, issue descriptions, and fix guidance.
![Opportunity group modal listing affected URLs with issue details and Fix with AI actions]()
Use this modal to:
* Confirm exactly which URLs are affected.
* Read the issue explanation and `How to fix` guidance.
* Decide whether to handle manually or trigger AI assistance.
## Step 3: Click `Fix with AI` for execution-ready help
Inside the modal, each row has a `Fix with AI` action. Clicking it opens AI help with the selected issue context so your team can move from diagnosis to draft fixes faster.
![Fix with AI flow open from opportunities context with assistant panel]()
Use `Fix with AI` when:
* The issue is clear but writing the fix takes time.
* You want a first draft for titles, descriptions, or content updates.
* You need consistent fix suggestions across many similar rows.
## Quick weekly checklist
1. Review groups by impact and effort.
2. Open the top 2 to 3 groups with the highest upside.
3. Validate URLs in the modal and assign owners.
4. Track unresolved-to-resolved movement.
5. Use `Fix with AI` for repetitive fixes to increase throughput.
## How to use filters
* `Impact = High`
Build a focused queue of groups most likely to move outcomes.
* `All impact levels`
Use this during planning, then narrow once you pick owners.
Use broad filters for planning, then narrow filters for execution handoff.
## What to fix first
| Pattern in groups table | What it usually means | Recommended action |
| ----------------------------------------------- | ------------------------------------- | --------------------------------------------------- |
| High impact + low effort group | Fast win with meaningful upside | Execute immediately |
| High action item count in one technical group | One root issue is affecting many URLs | Treat as a core technical task this sprint |
| Repeated content visibility issue across groups | Topic or template strategy is weak | Run one coordinated content update plan |
| High effort + medium impact | Useful but not urgent | Schedule after high-impact wins |
| Old last-checked date on high-impact group | Priority may be stale or unvalidated | Re-check data before assigning large execution work |
## Team routine
1. Weekly: choose priority groups and assign owners.
2. Bi-weekly: review group-level progress and reopen blocked items.
3. Monthly: compare technical vs content workstream impact.
4. Quarterly: remove recurring issue themes with systemic fixes.
## Keep in mind
* Group counts help prioritize, but the modal confirms exact URL-level work.
* `Fix with AI` accelerates drafting, but your team should still review before publishing.
* Always align group prioritization with business-impact pages.
## Where to go next
* [Opportunities](/data/opportunities/overview): prioritize URL-level execution
* [Technical SEO audit](/data/technical/seo-audit): investigate technical root causes in depth
* [Google Search keywords](/data/google-search/keywords): turn keyword-level gaps into updates
# Pages
Source: https://docs.atomicagi.com/data/opportunities/overview
Prioritize page-level SEO actions by impact, page value, and effort
![Opportunities pages overview with cards and page table]()
Use this page to choose which URLs your team should work on first, then move straight into issue-level fixes.
Important: Prioritize by business impact, not by issue
count alone.
## Questions this page should answer
1. Which pages have the highest-value SEO opportunities right now?
2. Which page should we open first for issue details?
3. When should we click `Fix with AI` instead of handling manually?
## Before you analyze
* Keep the same project and time context you use across Opportunities.
* Start with `All impact levels`, then narrow only after you see the full pattern.
* Prioritize `Page value` and `Impact` together before assigning work.
## What this page gives you
* Workflow cards: `All items`, `Unresolved`, `In review`, `Resolved`.
* A page table with `Page value`, `Impact`, and an action button per row.
* One-click drill-down into issue details for a specific URL.
* A direct `Fix with AI` action when you want execution help fast.
## How to read the top cards
* `All items`: total opportunity count in scope.
* `Unresolved`: items that still need work.
* `In review`: items being validated.
* `Resolved`: items marked complete.
Use these as workflow status signals, then prioritize in the table below.
Key signal: High page value with high impact should always
outrank high issue count on low-value pages.
## How these metrics are calculated (simple)
### Page value
Page value is a weighted score combining page business importance and performance potential.
### Impact
Impact is the estimated upside if that issue is resolved on the page (`Low`, `Medium`, `High`).
### Workflow status cards
Workflow status cards are counts of issue items by state (`All items`, `Unresolved`, `In review`, `Resolved`).
## Step 1: Start from the pages overview table
Use the table to pick one URL to open first.
* `Page value` tells you business importance.
* `Impact` tells you expected SEO upside.
* `Fix with AI` lets you jump directly to assisted execution for that row.
Start with rows where both `Page value` and `Impact` are high.
## Step 2: Open a page row to view issue-level details
Click a page to open its detail modal. This is where you see the exact issue, why it matters, and what to change.
![Page opportunities modal with issue details, how-to-fix guidance, and action button]()
Use this modal to:
* Validate the issue before assigning work.
* Read the built-in `How to fix` guidance.
* Decide if the fix should be manual or AI-assisted.
## Step 3: Click `Fix with AI` to generate execution help
After you click `Fix with AI`, the assistant opens with the selected URL and issue context already loaded.
![Opportunities pages with Fix with AI flow open in the assistant panel]()
Use `Fix with AI` when:
* The issue is clear and you need a fast first draft.
* You are handling repetitive fixes across many pages.
* You want consistent fix output for your team.
## Quick weekly checklist
1. Review high-value, high-impact rows first.
2. Open each top row and validate issue details in the modal.
3. Assign owner and due date per selected URL.
4. Use `Fix with AI` for repetitive or high-volume fixes.
5. Track unresolved-to-resolved movement weekly.
## How to use filters
* `Impact = High`
Build a focused queue for this sprint.
* `Page = specific URL`
Run a full issue check for one key page before publishing.
Start broad, then narrow once owners and priorities are clear.
## What to fix first
| Pattern in page table | What it usually means | Recommended action |
| ------------------------------------- | ---------------------------------------------- | ----------------------------------------- |
| High page value + high impact | Strong upside with clear business relevance | Prioritize this sprint |
| Many issues on one important page | One URL has stacked blockers | Bundle into one focused update cycle |
| High impact but lower page value | SEO upside exists but business return is lower | Schedule after core business pages |
| Low impact on low-value pages | Limited near-term return | Defer unless needed for technical hygiene |
| Same issues stay unresolved each week | Work is not getting completed or validated | Assign ownership and add weekly follow-up |
## Team routine
1. Weekly: pick priority URLs and assign work.
2. Bi-weekly: validate completed fixes and status movement.
3. Monthly: rebalance page priorities by business value and results.
4. Quarterly: remove recurring patterns with systemic fixes.
## Keep in mind
* Opportunity count alone is not priority; always pair with value and impact.
* AI suggestions still need human review before publishing.
* Very short windows can overreact to normal URL-level volatility.
## Where to go next
* [Opportunities groups](/data/opportunities/groups): prioritize by issue type and theme
* [Technical overview](/data/technical/overview): validate technical blockers behind page issues
* [Google Search landing pages](/data/google-search/landing-pages): compare opportunity priority with traffic behavior
# Reports
Source: https://docs.atomicagi.com/data/reports/overview
Build reusable report views from live project data and publish shareable links.
![Reports library page with predefined and custom report cards]()
Use this page to create, organize, and deliver reporting views for clients and internal teams.
Important: Treat reports as a reusable system. Keep
structure stable, update data and commentary each cycle.
## Questions this page should answer
1. Which reports already exist for this project?
2. Should I start from scratch or use a template?
3. How do I publish and share a report URL?
## Before you build reports
* Confirm you are in the correct project.
* Check data-source setup first so report blocks can load real data.
* Use naming conventions your team can scan quickly.
## What this page gives you
* A report library with two predefined cards (`Brand keywords`, `Questions`).
* Custom report cards created by your team.
* `Create report` action to open the template chooser.
* Card-level actions for editable reports (edit, duplicate, delete).
## Create report flow
![Create report modal with start-from-scratch and template options]()
When you click `Create report`, you can:
* `Start from scratch`: empty layout, add blocks manually.
* Choose a template: starts with predefined layout blocks.
**Your task:** Start with a template unless you have a clear custom layout requirement.
Current templates include patterns like `Warnings`, `Victories`, `SEO Overview`, `Conversion tracking`, `Page performance`, and `Geographic analysis`.
## Open and read a report
![Loaded report view with live table data and publish action]()
In report view you can:
* Read report blocks that render live analytics components.
* Use top actions to edit, duplicate, delete, or publish (if you have write access).
* Keep report structure stable while data refreshes from connected sources.
## Edit mode workflow
![Report edit mode with editable block title, description, and layout controls]()
In report edit mode, you can:
* Rename report title and description.
* Toggle block reorder mode.
* Add layout blocks by selecting:
data source -> section -> view type -> filters -> date range.
* Remove or reorder blocks.
Changes persist to the report configuration so the same layout can be reused in future periods.
## Publish and share
`Publish` exposes a shareable public URL for the report.
**Your task:** Publish only after final QA of filters, date range, and block titles.
Recommended workflow:
1. Finalize report title, description, and blocks.
2. Click `Publish`.
3. Copy the generated public URL.
4. Share with stakeholders.
## Quick weekly checklist
1. Remove stale drafts and duplicate cards.
2. Reuse templates for recurring reporting cycles.
3. Verify critical blocks still load expected data.
4. Re-publish after major report structure changes.
## Keep in mind
* Report quality depends on upstream data-source health.
* Predefined reports are fast shortcuts; custom reports give full layout control.
* Publish links are easiest to consume for non-product viewers.
## Where to go next
* [Google Search Overview](/data/google-search/overview)
* [AI Search Overview](/data/ai-search/overview)
* [Attribution](/data/attribution/overview)
# Cannibalization
Source: https://docs.atomicagi.com/data/technical/cannibalization
Find semantically overlapping pages so you can consolidate intent and reduce ranking conflicts
![Cannibalization table with page pairs, semantic similarity, and overlap match counts]()
Use this page to detect pages that compete for the same intent based on project knowledge similarity.
Important: A flagged pair is a prioritization signal, not
an automatic merge decision.
## Questions this page should answer
1. Which page pairs overlap too much in intent?
2. How severe is the overlap based on similarity and matching signals?
3. Which pairs should we merge, re-scope, or relink first?
## Before you analyze
* Make sure [Project Knowledge](/settings/project/knowledge) is synced recently.
* Start with high-value page types (money pages, category hubs, key guides).
* Review the top summary cards before reading pair-level rows.
## What this page gives you
* Summary cards:
* `Analyzed pages`
* `Detected pairs`
* `Similarity threshold`
* `Min overlap matches`
* Pair table with:
* `Page A`
* `Page B`
* `Similarity`
* `Overlap Matches`
## How to read the top cards
* `Analyzed pages`: how many indexed pages were checked.
* `Detected pairs`: number of overlapping pairs found.
* `Similarity threshold`: minimum semantic similarity required to flag a pair.
* `Min overlap matches`: minimum shared matches required before a pair appears.
Use this interpretation:
* High detected-pair count with stable page volume means structural overlap debt.
* Lower threshold values increase recall but add noisier pairs.
* Higher minimum overlap values reduce noise but can hide borderline conflicts.
## How to read the pair table
`Page A` and `Page B` are the two pages that may compete for the same intent.
* `Similarity`: semantic closeness score shown as a percent.
* `Overlap Matches`: number of overlapping signals found between the pair.
Prioritize pairs where:
* Similarity is high.
* Overlap matches are high.
* Both URLs target high-value funnel stages.
## What to do with flagged pairs
Use one of these actions for each pair:
1. Merge pages when both target the same core intent.
2. Split intent clearly by rewriting angle, scope, and title.
3. Strengthen canonical internal linking when both pages should exist.
4. Update metadata and headings to reduce ambiguity.
## Quick weekly checklist
1. Review highest-similarity pairs first.
2. Decide merge, split, or relink per pair.
3. Add implementation tickets with page owners.
4. Recheck overlap after updates are published and knowledge is synced.
## What to fix first
| Pattern in cannibalization table | What it usually means | Recommended action |
| ------------------------------------------- | ---------------------------------------- | -------------------------------------------- |
| High similarity and high overlap matches | Clear intent conflict | Merge or re-scope one of the pages |
| High similarity on commercial URLs | Revenue-impacting competition | Resolve this pair in the current sprint |
| Many pairs in one folder | Topic architecture is too fragmented | Consolidate cluster structure and linking |
| Repeated pairs after previous fixes | Knowledge sync or implementation gap | Re-sync knowledge and validate shipped edits |
| Low overlap but still high similarity score | Borderline conflict or broad topic scope | Tighten angle and search intent per page |
## Team routine
1. Weekly: triage new high-risk pairs.
2. Bi-weekly: verify shipped fixes reduced overlap.
3. Monthly: review recurring conflict clusters and template causes.
## Keep in mind
* This view depends on indexed project knowledge quality.
* Not every overlap is harmful if user intent and funnel stage are distinct.
* Fixes should be validated in rankings and conversion behavior, not only similarity scores.
## Where to go next
* [Project Knowledge](/settings/project/knowledge)
* [Interlinking](/data/technical/interlinking)
* [URL indexing](/data/technical/url-indexing)
* [SEO audit](/data/technical/seo-audit)
# Interlinking
Source: https://docs.atomicagi.com/data/technical/interlinking
Strengthen internal link structure to improve discoverability and authority flow
![Interlinking page with internal links table, score, incoming links, and outgoing links]()
Use Interlinking to find pages that are under-supported by internal links, then fix structure before rankings and crawl efficiency decline.
## Questions this page should answer
1. Which pages have weak internal-link support?
2. Which important pages are not receiving enough incoming links?
3. Where should we add internal links first to improve discoverability?
## Before you analyze
* Start with high-value page groups (commercial and core category pages).
* Sort by `Score` and check low-score pages first.
* Check both views (`Table` and `Mind map`) before final prioritization.
## What this page gives you
* Two views of the same internal-link dataset:
* `Table` view for page-level prioritization
* `Mind map` view for structure and cluster analysis
* URL-level metrics:
* `Score`
* `Incoming links`
* `Outgoing links`
* Clickable nodes in mind map to inspect incoming and outgoing relationships.
## How the Interlinking page is organized
The top-right view toggle switches between:
* `Table` icon: row-level internal-link health
* `Mind map` icon: network-style structure view
Use both in sequence:
1. Start in `Table` to identify weak pages.
2. Switch to `Mind map` to see why those pages are weak structurally.
3. Build fix tickets based on both row metrics and graph context.
## Table view: prioritize pages by link health
This is your execution view for weekly link fixes.
What each column means:
* `Score`: link health indicator for the page.
* `Incoming links`: how many internal pages support this URL.
* `Outgoing links`: how many internal paths this page gives to others.
How to interpret:
* Low score + very low incoming links means discoverability risk.
* High outgoing with low incoming can mean a page gives value but receives little support.
* Many pages with similar low score in one folder usually means a structural linking gap.
How to decide from table patterns:
* Low `Score` + low `Incoming links`:
* page is likely under-discoverable and under-supported.
* High `Incoming links` + low `Outgoing links`:
* page receives equity but does not distribute it.
* Good `Score` but weak commercial impact:
* check link source relevance, not only count.
## Mind map view: diagnose structure, clusters, and dead ends
![Interlinking mind map view showing internal URL nodes with incoming and outgoing link counts]()
Use this view to understand site structure at a glance.
What you are seeing:
* Each card is a URL node.
* `In:` is incoming internal links to that URL.
* `Out:` is outgoing links from that URL.
* Connecting lines show internal-link paths between levels.
* Node color reflects relative support strength (more incoming links = stronger node).
How to use the mind map controls:
* Drag to pan across the network.
* Use `Show more / Show less` controls per level to expand dense layers.
* Click any node to open detailed incoming/outgoing relationship lists.
How to interpret map patterns:
* Isolated nodes:
* usually orphan-like pages with weak internal distribution.
* Tight clusters with few bridges:
* strong local linking but weak cross-cluster authority flow.
* One dominant hub with many dependents:
* concentration risk if hub weakens or is de-prioritized.
## Practical workflow: table first, mind map second
Use this order:
1. Fix low-score commercial pages.
2. Fix low-score category and hub pages.
3. Switch to mind map and verify each priority page has strong path support.
4. Add links from high-authority hubs to isolated or weak nodes.
5. Re-check table scores after publishing.
## Quick weekly checklist
1. Sort by lowest `Score`.
2. Identify pages with weak `Incoming links`.
3. Validate weak pages in `Mind map` to confirm structural gaps.
4. Add contextual links from stronger, relevant pages.
5. Verify anchor text quality and topical match.
6. Recheck both views after updates are published.
## What to fix first
| Pattern in interlinking table | What it usually means | Recommended action |
| ---------------------------------------------- | ---------------------------------- | ------------------------------------------------------------ |
| Low score and near-zero incoming links | Page is isolated in site structure | Add links from relevant high-authority internal pages |
| Strong page with weak outgoing links | Link equity is not redistributed | Add contextual links to related commercial/content pages |
| Whole folder with weak scores | Template or navigation gap | Add section-level link modules and related-content blocks |
| Incoming links concentrated from one page type | Fragile link distribution | Diversify linking sources across templates |
| Node looks isolated in mind map | Page is structurally disconnected | Add bridge links from relevant hubs and nearby clusters |
| Dense cluster but weak cross-cluster links | Authority stays trapped locally | Add cross-cluster contextual links with strong anchor intent |
## Team routine
1. Weekly:
Review low-score rows and publish link fixes.
2. Bi-weekly:
Review mind map cluster health and bridge coverage.
3. Monthly:
Track score lift, incoming-link distribution, and affected traffic.
## Keep in mind
* Raw link count is not enough. Relevance and context matter most.
* Template-wide/footer links do not replace contextual in-content links.
* Interlinking results are cumulative; judge progress over several weeks.
* Use this page with `Cannibalization`, `Landing pages`, and `SEO audit` to align fixes with business impact.
## Where to go next
* [Cannibalization](/data/technical/cannibalization)
* [SEO audit](/data/technical/seo-audit)
* [URL indexing](/data/technical/url-indexing)
* [Google Search landing pages](/data/google-search/landing-pages)
# LLM audit
Source: https://docs.atomicagi.com/data/technical/llm-audit
Run a full AI-readiness audit and turn findings into a clear technical backlog
![LLM audit page with AI-readiness score cards and audit history table]()
Use LLM audit to check if your site is ready for AI discovery, then turn the findings into concrete fixes for engineering and SEO.
## Questions this page should answer
1. Is AI-readiness improving or stalling over time?
2. Which technical area is holding us back right now?
3. What should we fix first this sprint?
## Before you analyze
* Keep the same date range you use in AI Search pages.
* Compare the newest run with at least one prior run.
* Open the newest run detail before creating tickets.
## What this page gives you
* Five top scores: `Performance`, `Accessibility`, `Best practices`, `SEO`, `Content`
* `Audit history` so you can compare runs over time
* A full report preview with crawlability, schema, content structure, NLP, and Lighthouse-based diagnostics
## How to read the list view
* `Performance`: speed and technical execution quality
* `Accessibility`: structural clarity for users and machines
* `Best practices`: implementation hygiene and safety checks
* `SEO`: search-engine technical health
* `Content`: clarity and structure of on-page content for model understanding
How to interpret the top row:
* Low `Content` with high `SEO` usually means classic SEO is fine, but model parsing quality is weak.
* Low `Performance` can reduce crawl reliability and increase processing friction.
* Flat scores for months usually mean no active technical improvement cycle.
## How these scores are calculated (simple)
### Category score (`Performance`, `Accessibility`, `Best practices`, `SEO`, `Content`)
Category scores are weighted pass rates of checks in that category, shown on a `0-100` scale.
Higher score means fewer critical and major failures in that category.
## How to use audit history
Use `Audit history` as your change log:
1. Open the latest completed row.
2. Compare against the previous row.
3. Validate which fixes moved scores and which had no effect.
## LLM audit result preview: read it in this order
Start from the top and move block-by-block. This keeps the analysis focused and prevents random fixing.
## 1) LLM performance and crawlability
![LLM audit details top section with performance cards and crawlability checks]()
This top block tells you if AI systems can access and process your site reliably.
What to read first:
* `LLM performance` cards:
* `Performance`
* `Accessibility`
* `Best practices`
* `SEO`
* `LLM crawlability` status checks:
* `llms.txt status`
* `robots.txt status`
* `Sitemap status`
* `LLM Bots in robots.txt` allow/deny table.
What to do:
* If `robots.txt` or `sitemap` is not valid, fix that first.
* If important bots are blocked, adjust rules before content improvements.
* If score cards are weak and crawlability is healthy, move to deeper content and diagnostics sections.
## 2) Entity trust signals and schema validation
![LLM audit details with entity trust signals, schema validation, and recommendations]()
This section evaluates whether your site presents strong, machine-readable trust signals.
What’s included:
* `Entity trust signals` score.
* `Schema validation` checks, including:
* `HowTo`
* `Article`
* `FAQPage`
* `Organization`
* `BreadcrumbList`
* A recommendation list for missing or weak schema areas.
What to do:
* Add missing high-impact schema first (`Organization`, `Article`, `BreadcrumbList`, `FAQPage` where relevant).
* Keep schema aligned with actual page content.
* Use recommendation bullets as implementation tickets for dev/content teams.
## 3) Content structure and semantic coverage
![LLM audit details with content structure analysis, readability, NLP analysis, and HTML preview]()
This section explains whether your content is easy for models to understand.
What’s included:
* `Content score`
* `Readability`
* `Entity keyword coverage`
* `Readability grade level`
* `Content recommendations`
* `NLP analysis` terms and topic distribution
* Start of `Initial HTML preview`
How to interpret:
* Low readability with good keyword coverage means your content has topics, but clarity is weak.
* Weak entity coverage means key topics/entities are not explicit enough.
* NLP clusters show what your page is actually about from a machine perspective.
What to do:
* Simplify language in key sections.
* Improve heading hierarchy (`H2`/`H3`) and paragraph structure.
* Ensure important entities and terms appear naturally in primary sections.
## 4) HTML preview, performance tabs, and diagnostics
![LLM audit details with initial HTML preview and Lighthouse performance diagnostics]()
This block helps you verify what the crawler sees and where speed is lost.
What’s included:
* `Initial HTML preview` for source-level inspection.
* Score panel with tab views:
* `All`
* `FCP`
* `LCP`
* `TBT`
* `CLS`
* `Opportunities` with estimated savings.
* `Diagnostics` with implementation details.
What to do:
* Start with the highest estimated savings items.
* Fix render-blocking CSS/JS issues first.
* Track if performance changes improve both UX and crawl efficiency.
Quick definitions:
* `FCP` (First Contentful Paint): time until first visible content appears.
* `LCP` (Largest Contentful Paint): time until the main content block appears.
* `TBT` (Total Blocking Time): how long scripts block user interaction.
* `CLS` (Cumulative Layout Shift): visual instability while the page loads.
## 5) Accessibility and best-practices findings
![LLM audit details accessibility and best practices checks with contrast and labels findings]()
This section highlights structural quality and implementation hygiene.
What’s included:
* `Accessibility` checks (for example `Contrast`, `Names and labels`).
* `Best practices` checks:
* `Trust and Safety`
* `General`
What to do:
* Resolve contrast and labeling issues that block readability and machine interpretation.
* Fix recurring best-practice warnings to reduce technical fragility.
* Re-run audits after changes to confirm warning reduction.
## 6) SEO, content, and crawling/indexing checks
![LLM audit details with SEO, content, and crawling and indexing check sections]()
This section is your final validation layer before closing a run.
What’s included:
* `SEO` score and related checks.
* `Content` checks (title, description, and other core signals).
* `Crawling and indexing` checks (status/response-level validations).
What to do:
* Confirm core SEO and metadata checks pass on important pages.
* Fix crawl/index warnings before scaling content output.
* Use this section to verify technical readiness after implementation.
## If you see this, do this next
| What you see in the report | What it usually means | What to do next |
| ---------------------------------------- | ------------------------------------------------- | ------------------------------------------------------------------- |
| `robots.txt` or `Sitemap` is not valid | AI crawlers cannot access or trust your crawl map | Fix crawl directives and sitemap first |
| Entity trust score is low | Important trust schema is missing | Add `Organization`, `Article`, and `BreadcrumbList` schema first |
| Readability is low | Content is hard to parse and summarize | Simplify language and improve section structure |
| `LCP`/`TBT` problems in performance tabs | Rendering path is heavy | Remove blocking CSS/JS and optimize loading order |
| Accessibility warnings stay high | Structural quality debt remains | Fix labels, contrast, and semantic structure before scale |
| SEO/content checks fail | Core metadata/indexability issues remain | Resolve title/description/index checks before publishing more pages |
## Quick weekly checklist
1. Review list-view score direction.
2. Open latest audit details and verify crawlability first.
3. Prioritize entity/schema and content-structure gaps.
4. Fix top Lighthouse opportunities and accessibility warnings.
5. Validate SEO/content/crawling checks before marking complete.
## What to fix first
| Pattern in LLM audit | What it usually means | Recommended action |
| ---------------------------------------------- | ------------------------------------------------- | ----------------------------------------------------- |
| Crawlability status fails (`robots`/`sitemap`) | AI engines cannot access content correctly | Fix access rules and sitemap integrity first |
| Entity trust score is low | Structured trust signals are missing | Add/fix Organization, Article, FAQ, Breadcrumb schema |
| Readability low, entity coverage weak | Content is hard to parse semantically | Simplify language and strengthen entity-rich sections |
| Performance opportunities remain high | Technical performance debt affects render quality | Implement highest-savings fixes first |
| SEO/content checks pass but visibility is weak | Off-page and prompt-level coverage gap | Pair fixes with AI Search prompt and citation work |
## Team routine
1. Weekly: run list review + one detailed audit deep dive.
2. Bi-weekly: validate whether shipped fixes changed detail checks.
3. Monthly: report recurring technical blockers and closure rate.
## Keep in mind
* One score alone is not enough; read section-level diagnostics.
* Not every warning has equal impact. Prioritize crawlability and trust signals first.
* High classic SEO quality does not automatically mean strong AI discoverability.
## Where to go next
* [AI Search visibility](/data/ai-search/visibility)
* [Technical overview](/data/technical/overview)
* [SEO audit](/data/technical/seo-audit)
# Technical Overview
Source: https://docs.atomicagi.com/data/technical/overview
See technical health across SEO, AI readiness, and indexing so you can prioritize fixes faster
![Technical overview with SEO audit, LLM audit, URL indexing cards, and action shortcuts]()
Use this page as your technical control center. It helps you decide where to focus first before opening deeper Technical tabs.
Important: Evaluate trend direction, not one isolated score
snapshot.
## Questions this page should answer
1. Is overall site health improving or getting worse?
2. Is the current risk mainly from SEO issues, AI-readiness gaps, or indexing problems?
3. Which technical workflow should the team run next?
## Before you analyze
* Keep the same date range you use in Search and Attribution reviews.
* Start with no assumptions and check all three blocks: `SEO audit`, `LLM audit`, and `URL indexing`.
* Compare current values with the prior period before assigning work.
## What this page gives you
* A top-level snapshot of your technical health across crawl quality, AI-readiness, and indexing.
* Fast access to the main execution views: `SEO audit`, `LLM audit`, `Interlinking`, `Cannibalization`, and `URL indexing`.
* Action shortcuts so teams can launch technical work directly from the overview.
## How to read the top cards
Read the overview blocks in this order:
1. `SEO audit`: crawl health and issue pressure (`Errors`, `Warnings`, `Notices`).
2. `LLM audit`: readiness for AI engines (`Performance`, `Accessibility`, `Best practices`, `SEO`, `Content`).
3. `URL indexing`: discoverability and indexing coverage (`All URLs`, `Indexed URLs`, `Discovered`, `Unknown to Google`).
Use this rule:
* If `Errors` are rising, start with `SEO audit`.
* If `Content` or `Performance` is weak in LLM metrics, move to `LLM audit`.
* If indexed coverage is low or unknown URLs increase, move to `URL indexing`.
Key signal: Pick one primary technical workflow per review
cycle. Parallel fixes across all three areas usually slow execution and hide
impact.
## How these metrics are calculated (simple)
### Site health
```text theme={null}
Site health = 100 - Weighted issue severity impact from current technical findings
```
### LLM category scores
LLM category scores are weighted pass rates for checks in each category, shown on a `0-100` scale.
### Indexing status cards
Status cards are direct counts of tracked URLs in each indexing state.
## SEO audit section
This block shows whether technical SEO issues are controlled or spreading.
What to watch:
* `Site health` trend direction.
* Spikes in `Errors`.
* Large `Warnings` volume that can become future errors.
If error pressure is growing, open [SEO audit](/data/technical/seo-audit) immediately.
## LLM audit section
This block tells you if your site is structurally ready for AI answer engines.
What to watch:
* `Content` and `Performance` scores first.
* Any category with repeated decline across audits.
* Gaps between classic SEO quality and AI-readiness.
If one category is consistently weak, continue in [LLM audit](/data/technical/llm-audit).
## URL indexing section
This block helps you catch discoverability problems before they hurt growth.
What to watch:
* Difference between `All URLs` and `Indexed URLs`.
* Any increase in `Unknown to Google`.
* Stalled discovery for new pages.
If indexed coverage is low for important pages, continue in [URL indexing](/data/technical/url-indexing).
## Quick weekly checklist
1. Check all three blocks for direction change.
2. Flag the single highest-risk area (SEO issues, LLM quality, or indexing).
3. Open the matching detailed page and create fix tickets.
4. Confirm owners and deadlines for critical technical fixes.
5. Recheck trend change after updates are shipped.
## What to fix first
| Pattern on overview | What it usually means | Recommended action |
| ------------------------------------ | -------------------------------------------- | -------------------------------------------------- |
| Errors rising in SEO audit | Crawl quality is degrading | Open SEO audit and fix critical issues first |
| Low LLM content/performance | AI engines may struggle with content quality | Prioritize LLM audit recommendations for key pages |
| Indexed URLs lag far behind all URLs | Discoverability/indexing bottleneck | Run URL indexing checks and submit important pages |
| Multiple sections decline together | Technical debt is systemic | Run cross-functional technical sprint |
## Team routine
1. Weekly: monitor technical direction and assign highest-risk fixes.
2. Bi-weekly: review progress in each detailed Technical page.
3. Monthly: report resolved issues and remaining structural risks.
## Keep in mind
* A high health score can still hide critical issues in a few key pages.
* AI-readiness and indexing are not automatic by-products of classic SEO.
* Trends matter more than one-day values.
## Where to go next
* [SEO audit](/data/technical/seo-audit)
* [LLM audit](/data/technical/llm-audit)
* [Interlinking](/data/technical/interlinking)
* [Cannibalization](/data/technical/cannibalization)
* [URL indexing](/data/technical/url-indexing)
# SEO audit
Source: https://docs.atomicagi.com/data/technical/seo-audit
Run end-to-end technical SEO triage from audit summary to URL-level fixes in one workflow
![SEO audit overview with site health cards and audit history table]()
Use this page as your full technical SEO workflow. It covers the summary view, run-level diagnosis, crawl-log validation, and URL-level issue execution.
Important: Fix highest-severity template-level issues
first. They usually unlock the fastest site-wide improvement.
## Questions this page should answer
1. Is technical health improving from audit to audit?
2. Which issue groups are causing the biggest risk right now?
3. Which exact URLs should be fixed first this sprint?
## Before you analyze
* Keep the same date range used in your weekly reporting cadence.
* Start with the summary cards, then open the newest completed run.
* Compare at least two runs before assigning root cause.
## What this page gives you
* Top-level technical health snapshot (`Site health`, `Errors`, `Warnings`, `Notices`).
* Run-level diagnosis with issue grouping and crawl behavior.
* URL-level issue detail to create precise execution tasks.
* A complete triage path in one document.
## How to read the top cards
* `Site health`: overall technical quality score for the crawl.
* `Errors`: highest-priority issues that can block crawling/indexing quality.
* `Warnings`: medium-priority issues that can create risk if ignored.
* `Notices`: lower-priority issues to batch and resolve over time.
Use this interpretation:
* Rising `Errors` means immediate action.
* Flat but high `Warnings` means hidden debt that can become future blockers.
* Stable `Site health` with concentrated issues still requires focused fixes.
Key signal: If errors rise and pages crawled stay stable,
the issue is usually real quality degradation, not sampling noise.
## How these metrics are calculated (simple)
### Site health
```text theme={null}
Site health = 100 - Weighted impact of all detected issues in the selected audit run
```
Higher score means fewer severe issues across crawled pages.
### Errors / Warnings / Notices
Severity counts are direct totals of findings grouped by severity (`Error`, `Warning`, `Notice`).
## How to use audit history
Use `Audit history` to choose the run to investigate first:
* Open the newest completed run first.
* Compare with the previous run for issue-trend confirmation.
* Watch `Pages crawled` and `Duration` to detect crawl volatility.
If a run looks abnormal, continue with run-level diagnosis below.
## Run detail: All issues view
This is where you prioritize issue categories by impact before touching individual URLs.
![SEO audit run detail All issues view with category-level issue analysis]()
What to do:
* Identify issue groups affecting the most pages.
* Prioritize error classes on key templates first.
* Choose one high-impact issue class and close it end-to-end.
What this tells you:
* Whether risk is concentrated in one technical area.
* Whether the issue is likely template-wide or page-specific.
## Run detail: Crawl log view
Use crawl log when you need evidence about how URLs were crawled and where failures happened.
![SEO audit run Crawl log view with per-URL crawl results and statuses]()
What to check:
* Repeated HTTP errors, timeouts, or blocked responses.
* Failure clusters by folder/template.
* Crawl behavior changes after deployments.
When to use:
* Issue counts changed sharply.
* Developers need URL-level crawl proof.
## If the crawl log shows 403 responses
A `403` means the website, firewall, or CDN blocked the crawler from reading the page. If the site uses Cloudflare or another bot protection tool, allow the Atomic AI site audit crawler by user agent:
```text theme={null}
AtomicAGIBot/1.0
```
Atomic AI does not provide fixed IP addresses for site audit allowlisting. Use the crawler user agent instead of an IP allowlist.
Allow this user agent for the whole domain. The audit needs domain-wide access so it can crawl the homepage, `robots.txt`, sitemap files, and all public pages included in the audit.
After updating the firewall or bot rules, run the audit again and check the crawl log for successful `2xx` or expected redirect responses.
## Issue detail: affected URLs view
After selecting an issue class, use issue detail to assign exact fixes.
![SEO issue detail view with affected URLs for one technical issue]()
How to work this table:
1. Group URLs by template/folder.
2. Prioritize commercial and high-visibility pages.
3. Assign owner and deadline per URL group.
4. Validate fixes in the next audit run.
Use this interpretation:
* Many URLs in one path usually means a shared template defect.
* Mixed status behavior can indicate rendering/routing inconsistency.
## Quick weekly checklist
1. Check summary cards for direction change.
2. Open latest run and triage top issue groups.
3. Use crawl log to confirm root cause.
4. Move to issue-detail URL lists and assign work.
5. Validate reduction in the next completed run.
## What to fix first
| Pattern in SEO audit flow | What it usually means | Recommended action |
| ---------------------------------------- | --------------------------------------------- | --------------------------------------------------------- |
| Errors up sharply in summary | New release or template issue introduced risk | Open latest run and prioritize critical issue class |
| One issue group affects many URLs | Template-level defect | Patch shared template/component first |
| Crawl log shows repeated failures | Crawl access/infrastructure instability | Fix technical crawl blockers before content tweaks |
| Same issue persists across runs | QA/deployment process gap | Add pre-release technical checks and owner accountability |
| High-value URLs affected in issue detail | Direct business impact | Fix those URLs first in current sprint |
## Team routine
1. Weekly: run full triage from summary to issue-detail assignments.
2. Bi-weekly: verify fix closure and rerun checks.
3. Monthly: report resolved issue classes and repeated root causes.
## Keep in mind
* One clean run does not prove long-term stability.
* Crawl-size shifts can change totals without real quality movement.
* Never mark fixes complete before the next successful audit confirmation.
## Where to go next
* [Technical overview](/data/technical/overview)
* [LLM audit](/data/technical/llm-audit)
* [Interlinking](/data/technical/interlinking)
* [Cannibalization](/data/technical/cannibalization)
* [URL indexing](/data/technical/url-indexing)
# URL indexing
Source: https://docs.atomicagi.com/data/technical/url-indexing
Track index status and resolve indexing issues before they impact growth
![URL indexing page with URL submission input, indexing status cards, and URL status table]()
Use this page to monitor index coverage and resolve blocked or failed URLs before visibility drops.
## Questions this page should answer
1. Are key pages being indexed correctly?
2. Which URLs are failing indexing and why?
3. Which indexing actions should happen first this week?
## Before you analyze
* Start with high-priority page groups (money pages and key content hubs).
* Review status counts before checking individual URLs.
* Separate `Errored` pages from pages that are simply not yet discovered.
## What this page gives you
* URL submission tools (`Check Status`, `Request Indexing`).
* Indexing status cards:
* `All URLs`
* `Indexed URLs`
* `Discovered`
* `Unknown to Google`
* `Indexing requested`
* `Errored`
* A URL-level status table with last-check information.
## How to read the status cards
* `All URLs`: tracked URLs in this project.
* `Indexed URLs`: pages currently known and indexed.
* `Discovered`: found but not yet fully indexed.
* `Unknown to Google`: not recognized in index checks.
* `Indexing requested`: URLs sent for indexing recently.
* `Errored`: URLs with indexing check failures.
How to interpret:
* High `Errored` count means direct indexing risk.
* Large gap between `All URLs` and `Indexed URLs` means coverage opportunity.
* Rising `Unknown to Google` often means discovery or sitemap issues.
## How to use the URL table
Use the table to create an execution queue:
* Sort and review failed URLs first.
* Group issues by path/template to find shared root causes.
* Check `Last checked` to avoid acting on stale assumptions.
Use bulk actions:
* `Load URLs from sitemap` to refresh candidate URLs.
* `Index all URLs` or targeted requests after fixes are live.
## Quick weekly checklist
1. Review status-card changes from last week.
2. Triage `Errored` URLs first.
3. Submit fixed priority URLs for re-indexing.
4. Validate status movement after recheck.
5. Escalate repeated failures to engineering.
## What to fix first
| Pattern in URL indexing | What it usually means | Recommended action |
| ------------------------------------ | ------------------------------------------ | -------------------------------------------------------- |
| High errored count | Technical/indexing blockers are unresolved | Fix root cause and re-request indexing for priority URLs |
| Large unknown-to-google segment | Discovery pathway is weak | Validate sitemap, internal links, and crawl access |
| Many discovered but not indexed URLs | Quality/signaling gap | Improve page quality and internal support links |
| Important page stuck unindexed | Direct business risk | Prioritize page-level fix and immediate re-index request |
## Team routine
1. Weekly: index-status review and top-priority URL triage.
2. Bi-weekly: investigate recurring failure patterns.
3. Monthly: report coverage growth and persistent indexing blockers.
## Keep in mind
* Indexing changes may take time after requests.
* Requesting indexing without fixing root cause rarely works.
* Priority should follow business impact, not URL count alone.
## Where to go next
* [Technical overview](/data/technical/overview)
* [SEO audit](/data/technical/seo-audit)
* [Cannibalization](/data/technical/cannibalization)
* [Interlinking](/data/technical/interlinking)
* [Google Search landing pages](/data/google-search/landing-pages)
# Welcome to Atomic AGI
Source: https://docs.atomicagi.com/introduction
Understand what Atomic AGI tracks and how to use these docs to move from insight to action
![Google Search overview in Atomic AGI with trend and metrics]()
Atomic AGI helps SEO teams monitor Google and AI search performance, find the highest-impact issues, and execute fixes faster.
## What this documentation gives you
* A clear path from setup to first outcomes
* Page-by-page guidance for every Data view
* Practical workflows for Automation and Settings
Use these docs when you need to answer:
1. What changed and why?
2. What should we prioritize this week?
3. What should we do next?
## What Atomic AGI tracks
* Google Search performance by keyword, page, location, and device
* AI search visibility, citations, prompts, and sentiment
* Technical issues that block discovery and indexing
* Opportunity queues to prioritize what to fix first
* Attribution and reporting views for stakeholder updates
## How to use these docs effectively
1. Open the same page in the app and in docs.
2. Read the "Questions this page should answer" section first.
3. Use the interpretation and checklist sections to decide actions.
4. Assign fixes, then validate status changes in the next cycle.
## Documentation map
* `Get started`
Orientation and first-run setup flow.
* `Data`
How to read analytics pages and decide priorities.
* `Automation`
How to run one-off analysis and recurring execution.
* `Settings`
How to manage profile, organization, and project configuration.
## Start here
Continue to the [Quickstart Guide](/quickstart) to set up your first project and generate your first prioritized action list.
# Quickstart Guide
Source: https://docs.atomicagi.com/quickstart
End-to-end setup from organization creation to GSC, GA4, and conversion events
Use this page to complete onboarding in the correct order. You will finish the initial setup first, then move to AI prompts
as the next step.
## What this setup gives you
* One new Atomic AGI account
* One organization (workspace) for your team
* One project connected to verified Google data
* Google Search Console + GA4 + conversion events configured
* Clean base for SEO, AI Search, and attribution reporting
## Step 1: Register your account
Start on the sign-up page and create your account.
![Registration form with email, full name, and password fields]()
Choose either:
* `Sign up with Google`
* Email + password sign-up
After successful registration, you will continue into organization setup.
## Step 2: Create your organization
An **organization** is your team workspace in Atomic AGI. It contains:
* Projects
* Members and permissions
* Shared billing and settings
![Join page with create organization form]()
Enter your organization name and click `Create organization`.
## Step 3: Enter your project details
![Set up your project form with the required Website URL field]()
After organization creation, you will move to project setup automatically.
Enter:
* `Company name`
* `Website URL`
* `Industry`
* `Target region`
The Website URL is required. Enter the full URL, including `http://` or `https://`, such as `https://example.com`.
Atomic saves only the website host as the project domain. Paths, query parameters, and fragments are removed, `www` is
normalized away, and the saved value always uses `https://`. For example, `http://www.example.com/blog?ref=setup` is
saved as `https://example.com`.
Select `Continue`. On the next page, continue to Google Search Console and use a Google account that has access to the
website you want to connect.
## Step 4: Select your Google Search Console property
![Google Search Console property selection list]()
Choose the GSC website you want to track. Atomic compares its hostname with the Website URL saved in Step 3.
* If the hostnames match, select `Continue`.
* If the hostnames differ, the button changes to `Connect and change domain`.
Atomic compares hostnames rather than complete URLs. For example, `https://www.example.com/page` and
`sc-domain:example.com` are treated as the same website.
### Confirm a different GSC domain
![Change project domain confirmation showing the current project domain and selected GSC domain]()
When the hostnames differ, Atomic shows `Change project domain?` before connecting the GSC website. Review both domains,
then choose one action:
* `Cancel` keeps the current project domain and returns you to the website picker without connecting the selected site.
* `Connect and change domain` connects the selected GSC website and replaces the project domain with that website.
Important: If the expected site is missing, verify it first
in Google Search Console, then reconnect.
## Step 5: Select your GA4 property
![GA4 property selection during onboarding]()
This GA4 step appears right after GSC selection. Choose the GA4 property for the same website/project and click `Next`.
## Step 6: Confirm the correct GA4 property and continue
![Storly.ai GA4 property selected with next button ready]()
The screenshot shows one example selection. In your workspace, select your own GA4 property.
Keep GSC and GA4 from the same business/project to avoid reporting mismatch.
## Step 7: Select your conversion event(s) and connect
![Conversion events step with book_demo_click selected]()
Choose the conversion event that represents a real business outcome in your project, then click `Connect`.
The screenshot shows `book_demo_click` as an example.
Examples:
* `book_demo_click`
* lead form submit
* trial start
## Initial setup complete: what should be connected
At this point, your onboarding setup is done when all three are connected:
* Google Search Console
* Google Analytics 4
* Conversion events
## Where to go next
AI prompts are the next phase after initial setup.
Next, go to [AI Search Prompts](/data/ai-search/prompts) and add the prompts you want to track.
![AI Search prompts page for setting tracked prompts]()
Use prompts across 3 intent layers:
1. Awareness: broad discovery questions.
2. Comparison: alternatives and evaluation questions.
3. Decision: high-buying-intent questions.
Continue with:
* [AI Search Prompts](/data/ai-search/prompts)
* [Project Data Sources](/settings/project/data-sources)
# Billing
Source: https://docs.atomicagi.com/settings/organization/billing
Review your workspace subscription, project plans, renewal, and member seats in one place.
![Workspace billing summary with one subscription, renewal details, seats, and project plans]()
Use Billing to understand the full recurring cost for your workspace. Every paid project plan and additional member seat appears on one subscription.
## Questions this page should answer
1. What is our recurring subtotal and next renewal date?
2. Which plan is assigned to each project?
3. Do we have enough seats for current members?
## Before you use this page
* Open Organization settings and select `Billing`.
* Confirm you have billing write access before changing plans or seats.
* Resolve a past-due invoice before making a billing change.
* On the first paid-plan checkout, Stripe collects the billing address and offers company Tax ID entry. Stripe saves
the supplied legal company name, address, and Tax ID to the workspace billing customer.
## What this page gives you
* One workspace subscription status and renewal date.
* One recurring subtotal for project plans and extra seats.
* The plan and price assigned to every project.
* Seats included by project plans, purchased seats, and seats currently in use.
* One `Company details` action for the legal company name, billing address, and Tax IDs stored in Stripe.
* One `Payment methods & invoices` action for payment methods, invoices, billing address, Tax ID, and cancellation.
## Change a project plan
Select `Change plan` beside a project. The dialog shows the available paid tiers, including the current plan.
* Starter and higher plans include MCP access.
* Team and higher plans include Grids.
* Managed is a custom, done-for-you option for teams that need higher limits, custom models, integrations, workflows, or AI Employees. Select `Book a demo` to discuss the setup with Atomic.
![Change plan dialog with paid project plan options]()
* Selecting Upgrade or Downgrade starts the appropriate action immediately; there is no separate Continue step.
* If a workspace trial already covers one paid project, adding a second paid project first asks for confirmation. Confirming ends the workspace trial and starts billing for all paid projects immediately.
* Upgrades are invoiced immediately.
* Downgrades create a credit for the next invoice.
* The workspace renewal date stays the same.
* A subscription scheduled to cancel cannot be changed or revived.
Atomic asks you to confirm the project name, old plan, new plan, when access changes, and the billing effect.
To cancel one paid project plan, open the `...` menu beside `Change plan` and select `Cancel plan`. Confirming moves only that project to Free and removes its recurring item. Other project plans and additional seats stay unchanged. If the project owns the final recurring item, its workspace subscription ends immediately without creating an invalid empty subscription or an immediate final invoice.
### Free trials cover one project
The first project on Starter or Team can start the 7-day workspace trial. Team Max starts as a paid plan without a
trial. While a workspace trial is active, other Free projects cannot select a paid plan.
![Upgrade dialog explaining that another project cannot become paid while the workspace trial is active]()
The paid-plan actions stay disabled until the workspace trial ends. Changing the tier of the original trial project remains available and does not end its trial.
## Change additional seats
![Change additional workspace seats dialog with quantity and recurring price]()
Select `Change seats`, enter the total number of separately purchased seats, and confirm the change.
* Adding seats is invoiced immediately.
* Removing seats creates a credit for the next invoice.
* You cannot remove seats that are required by current members.
* Additional seats remain part of the same workspace subscription.
## Manage payment and cancellation
Select `Company details` to save the legal company name, registered billing address, and primary Tax ID directly to
the workspace's Stripe customer. Atomic infers Stripe's required Tax ID type from the country of incorporation, so
customers only enter the Tax ID number in their country's valid format. Existing additional Tax IDs remain in Stripe.
![Company billing details dialog with legal company name, Tax ID, country, and registered billing address]()
Select `Payment methods & invoices` to open Stripe. You can update payment methods, download invoices, manage
additional Tax IDs, or manage the whole workspace subscription.
Whole-workspace cancellation affects every paid project. Use `Change plan` for an individual downgrade, or open the adjacent `...` menu and select `Cancel plan` to move one project to Free.
## Quick weekly checklist
1. Check subscription status and the next renewal date.
2. Compare seats in use with seats available.
3. Review project plans before adding a new project or member.
4. Resolve past-due invoices before requesting plan or seat changes.
## Keep in mind
* One-time additional AI-credit purchases are separate from the recurring subscription.
* Free projects do not add a zero-price item.
* If billing is scheduled to cancel, contact support before making a new commercial decision.
## Where to go next
* [Organization Members](/settings/organization/members)
* [Organization Projects](/settings/organization/projects)
* [Project Members](/settings/project/members)
# General
Source: https://docs.atomicagi.com/settings/organization/general
Review organization identity and perform high-impact organization actions.
![Organization general settings showing workspace name and delete workspace action]()
Use this page for top-level organization identity and lifecycle actions.
Important: Organization-level changes can affect every
project in the workspace. Confirm scope before using destructive actions.
## Questions this page should answer
1. Is the organization name correct?
2. Who can perform destructive organization actions?
3. Is this the right scope for the change I need?
## Before you use this page
* In the app, open Organization settings and select `General`.
* Confirm you have the required role for write actions.
* Double-check scope before destructive operations.
## What this page controls
* Organization/workspace name (displayed read-only in this view).
* `Delete workspace` action for admins.
* Entry point to other organization tabs (`Integrations`, `Members`, `Projects`, `Billing`).
## What this page gives you
* A fast identity check for the active organization.
* A clear boundary between organization-level settings and project-level settings.
* The destructive workspace action in one predictable location.
## How to use this page
### Treat delete as an irreversible action
`Delete workspace` removes workspace data, projects, members, and settings. Use only when intentionally decommissioning the organization.
### Use this page for identity checks
When auditing environment correctness, start by confirming the organization name in this tab.
### Move to specialized tabs for operations
Member management, integrations, and billing are handled in their dedicated organization tabs.
## Quick monthly checklist
1. Confirm organization identity is still correct.
2. Revalidate who has access to destructive organization actions.
3. Route operational tasks to the correct organization tab.
## What to fix first
| Pattern in Organization General | What it usually means | Recommended action |
| --------------------------------------------- | ------------------------------------ | ------------------------------------------ |
| Wrong organization name visible | You may be in the wrong workspace | Switch workspace before changing settings |
| Delete action is available to too many people | Permission scope is too broad | Review organization member permissions |
| User is looking for project setup | They are in the wrong settings scope | Move to the matching Project settings page |
## Team routine
1. Monthly: confirm organization identity and destructive-action ownership.
2. After team changes: review workspace-level permissions.
3. Before deletion: confirm projects, billing, data exports, and owners.
## Keep in mind
* Changes here can affect all projects in the organization.
* Most daily operations happen in `Members`, `Projects`, `Integrations`, and `Billing`.
## Where to go next
* [Project Data Sources](/settings/project/data-sources)
* [Organization Members](/settings/organization/members)
* [Organization Billing](/settings/organization/billing)
# Integrations
Source: https://docs.atomicagi.com/settings/organization/integrations
This organization route redirects to project integrations, where Slack and publishing connections are managed.
This organization route redirects to **Project settings -> Integrations**. Use it when you expected organization-level integrations but need the current project-level connection page instead.
Important: Slack and publishing connections are
project-scoped. Check the active project before editing credentials or
channels.
## Questions this page should answer
1. Why did the old organization integrations route move?
2. Where are Slack and publishing connections managed now?
3. Which project settings page should I open next?
## Before you continue
* Confirm you are in the correct project.
* Check whether the integration you need is project-specific.
* Use project-level settings for Slack channel, WordPress, and Webflow connections.
## What this page gives you
* A pointer from the old integrations location to the current project integrations page.
* A reminder that publishing connections are project-scoped.
* A safe path to the page that actually controls connection state.
Use the project integrations page to manage:
* Slack bot connection and channel selection for the current project.
* Reconnect flows when the project Slack app is missing the `app_mentions:read` scope.
* Publishing connections such as WordPress and Webflow.
## How to use this redirect page
Open [Project Integrations](/settings/project/integrations), then confirm the project name before connecting or editing anything.
## Quick setup checklist
1. Open the current project.
2. Go to [Project Integrations](/settings/project/integrations).
3. Confirm Slack and publishing connections belong to this project.
4. Verify each connection after editing credentials.
## What to fix first
| Pattern on this route | What it usually means | Recommended action |
| -------------------------------------- | ------------------------------------------------ | --------------------------------------------- |
| You expected organization integrations | Old navigation or notes are stale | Open Project Integrations |
| Slack destination looks wrong | Active project may be different | Switch to the intended project before editing |
| Publishing connection is missing | Destination is configured elsewhere or not added | Add it from Project Integrations |
## Team routine
1. When sharing setup docs: link to Project Integrations directly.
2. After project changes: verify Slack and publishing destinations again.
3. During onboarding: explain that publishing setup is project-scoped.
## Keep in mind
* Organization members and billing remain organization-level settings.
* Slack and publishing connections are managed per project.
* Old route names may still appear in older internal notes or screenshots.
## Where to go next
* [Project Data Sources](/settings/project/data-sources)
* [Project Integrations](/settings/project/integrations)
* [Organization Members](/settings/organization/members)
* [Organization Billing](/settings/organization/billing)
# Members
Source: https://docs.atomicagi.com/settings/organization/members
Invite members, configure role/scope, and manage feature-based access control.
![Organization members table with roles, project access, and action buttons]()
Use this page to manage organization access with role-based and feature-based permission control.
## Questions this page should answer
1. Who currently has access to this organization?
2. How do I invite a new member with the right scope?
3. How do `None`, `View only`, and `Full` actually affect access?
## Before you use this page
* In the app, open Organization settings and select `Members`.
* Confirm you have members write access.
* Check seat availability before bulk invites.
## What this page controls
* Member directory (`Email`, `Role`, `Project access`, `Joined at`, `Actions`).
* Invite flow with role + workspace permissions + project permissions.
* Edit permissions flow for existing members.
* Member removal.
## Invite members flow
![Invite members dialog showing email input, role selector, workspace permissions, and project access]()
`Invite members` opens a dialog where you define access before sending invites.
### How to invite correctly
1. Add one or multiple emails.
2. Select role:
`Admin`: full access across workspace and projects.
`Member`: granular permissions are applied.
3. Set `Workspace permissions` (`General`, `Integrations`, `Members`, `Projects`, `Billing`).
4. Select project access.
5. For each selected project, set feature/settings permissions.
## Edit permissions flow
![Edit permissions dialog with role, workspace permissions, and project feature permissions]()
Use the shield icon in the members table to open `Edit permissions` for an existing member.
### What you can edit
* Role (`Admin` or `Member`).
* Workspace permission scopes.
* Project access selection.
* Per-project feature + settings permission scopes.
## Permission scopes explained
The same 3 states are used across workspace and project permission rows:
| Scope | What it means | Typical use case |
| ----------- | ----------------------------------------------------------------------------- | -------------------------------------------- |
| `None` | Feature is hidden/inaccessible for that member in that scope | Restrict access entirely |
| `View only` | Member can see/read but cannot perform write actions (create/edit/delete/run) | Analyst/reviewer access without write rights |
| `Full` | Member can view and perform write actions | Owner/operator who executes work |
## Workspace-level permissions
These control access to organization settings sections:
* `General`: `/settings/organization/general`
* `Integrations`: project-level integration settings
* `Members`: `/settings/organization/members`
* `Projects`: `/settings/organization/projects`
* `Billing`: `/settings/organization/billing`
## Project feature permissions
These control feature access inside selected projects.
| Permission key | UI label | Docs link |
| ------------------ | -------------- | -------------------------------------------------------------------- |
| `GOOGLE_SEARCH` | Google search | [/data/google-search/overview](/data/google-search/overview) |
| `AI_SEARCH` | AI search | [/data/ai-search/overview](/data/ai-search/overview) |
| `ATTRIBUTION` | Attribution | [/data/attribution/overview](/data/attribution/overview) |
| `TECHNICAL` | Technical | [/data/technical/overview](/data/technical/overview) |
| `CONTENT_ANALYSIS` | Opportunities | [/data/opportunities/overview](/data/opportunities/overview) |
| `CONTENT_EDITOR` | Content editor | [/settings/project/brand-kit](/settings/project/brand-kit) |
| `CUSTOM_REPORTS` | Reports | [/data/reports/overview](/data/reports/overview) |
| `AI_AGENTS` | AI agents | [/automation/agents/overview](/automation/agents/overview) |
| `WORKFLOWS` | Workflows | [/automation/workflows/overview](/automation/workflows/overview) |
| `AUTOMATIONS` | Automations | [/automation/automations/overview](/automation/automations/overview) |
| `BRAND_KIT` | Knowledge | [/settings/project/brand-kit](/settings/project/brand-kit) |
## Project settings permissions
These control access to project settings tabs:
| Permission key | UI label | Docs link |
| ----------------------- | ------------ | ---------------------------------------------------------------- |
| `SETTINGS_GENERAL` | General | [/settings/project/general](/settings/project/general) |
| `SETTINGS_MEMBERS` | Members | [/settings/project/members](/settings/project/members) |
| `SETTINGS_DATA_SOURCES` | Data sources | [/settings/project/data-sources](/settings/project/data-sources) |
| `SETTINGS_PUBLISHING` | Integrations | [/settings/project/integrations](/settings/project/integrations) |
## Practical permission patterns
* Client reviewer: `View only` on reporting features, `None` on members/billing, limited project scope.
* SEO operator: `Full` on data/technical/workflows in assigned projects, `None` on billing.
* Organization admin: `Admin` role when they need global control.
## Quick weekly checklist
1. Resolve stale pending invites.
2. Audit members with broad `Full` scopes.
3. Remove access for users no longer on account.
4. Re-check project assignment breadth for each member.
## Keep in mind
* Organization members can have different scopes per project.
* Role and permission scopes interact: `Admin` bypasses granular restrictions.
* Seat limits can block new invites until capacity is increased.
## Where to go next
* [Organization Billing](/settings/organization/billing)
* [Organization Projects](/settings/organization/projects)
* [Project Members](/settings/project/members)
# Projects
Source: https://docs.atomicagi.com/settings/organization/projects
Review all organization projects and jump to project-level settings.
![Organization projects table showing project names, domains, and settings actions]()
Use this page to manage project inventory and navigate into individual project settings.
Important: Treat the project list as the workspace source
of truth. Duplicate or stale projects make reporting and permissions harder
to reason about.
## Questions this page should answer
1. Which projects are currently in this organization?
2. Which domains are linked to each project?
3. How do I open settings for a specific project quickly?
## Before you use this page
* In the app, open Organization settings and select `Projects`.
* Check if you have write permission for project creation.
* Validate domain ownership before creating new projects.
## What this page controls
* Project table (`Name`, `Domain`, `Actions`).
* `New project` creation action.
* `Settings` shortcut for each listed project.
## What this page gives you
* A workspace-level inventory of active projects.
* Domain context for each project.
* A shortcut into project-specific configuration.
## How to use this page
### Keep project inventory clean
Use this table as your canonical list of active organization projects.
### Create projects with intent
Use `New project` only when a distinct domain/use case warrants a separate configuration and reporting scope.
### Use row-level settings shortcuts
Use each row’s `Settings` button to jump directly into project configuration without manual navigation.
## Quick monthly checklist
1. Remove or archive stale projects operationally.
2. Audit duplicate domains and naming drift.
3. Validate each active project has an owner.
## What to fix first
| Pattern in Projects table | What it usually means | Recommended action |
| ---------------------------- | ---------------------------------- | --------------------------------------------------- |
| Duplicate or similar domains | Project scope may be unclear | Confirm which project owns the live domain |
| Project has no clear owner | Follow-through risk is high | Assign an owner before relying on reporting |
| Wrong project name/domain | Team may analyze the wrong surface | Open project settings and correct operational notes |
| Too many inactive projects | Workspace navigation becomes noisy | Retire or archive stale work operationally |
## Team routine
1. Monthly: review active projects and owners.
2. After new launches: confirm domain/project mapping.
3. Before audits: remove duplicate or stale project confusion.
## Keep in mind
* This page is about project inventory, not deep project configuration.
* Detailed project settings are managed in the `Project` group.
## Where to go next
* [Project General](/settings/project/general)
* [Project Data Sources](/settings/project/data-sources)
* [Project Integrations](/settings/project/integrations)
# Affiliate
Source: https://docs.atomicagi.com/settings/profile/affiliate
Understand the affiliate program, share your link correctly, and track referrals.
![Affiliate settings showing generated referral link and referrals list]()
Use this page to run your affiliate program activity in one place: generate your link, share it, and track who signs up through you.
Important: Always share the generated URL from this page.
Manually edited referral links are the most common attribution mistake.
## Questions this page should answer
1. Do I already have an affiliate link?
2. What is the affiliate program used for?
3. How do I copy and share the correct referral URL?
4. Which referrals are active vs inactive?
## Before you use this page
* In the app, open your profile settings and select `Affiliate`.
* If no code exists yet, generate one first.
* Use the exact generated URL when sharing.
## What this page controls
* Affiliate link generation (with terms confirmation on first generation).
* One-click copy for the referral URL.
* Referral list with user status badges (`Active` / `Inactive`).
## What the affiliate program is
Atomic AGI's affiliate program lets you invite new users with your unique referral link.
* When someone signs up using your link, they are attributed to your affiliate code.
* Referred users appear in your referrals list so you can track activity.
* Program terms and eligibility details are confirmed during affiliate code generation.
## How to use it end to end
1. Generate your affiliate code once.
2. Copy the exact referral URL from this page.
3. Share that URL in channels you own (email, social, website, communities).
4. Review referral status regularly and improve where conversion quality is low.
## How to use this page
### Generate your code once
If you do not have a code yet, click `Generate affiliate code`, review the terms dialog, and confirm.
### Share the exact link
Use the copy button so the full URL includes your affiliate parameter.
### Track referral quality
Review the referrals list regularly to see who converted into active subscribers and where follow-up is needed.
## Quick weekly checklist
1. Confirm your affiliate link still copies correctly.
2. Check new referrals and active/inactive mix.
3. Flag large changes in referral activation rate.
## What to fix first
| Pattern in Affiliate page | What it usually means | Recommended action |
| -------------------------- | -------------------------------------- | ------------------------------------------------ |
| Referral link is missing | Affiliate code was not generated | Generate the code and accept the terms |
| Signups are not attributed | Wrong or edited URL may have been used | Copy and share the exact generated link |
| Many inactive referrals | Traffic quality or onboarding is weak | Review source channel and follow-up expectations |
| Referral list is empty | No tracked signups yet | Test the copied link before broader sharing |
## Team routine
1. Weekly: review referral status changes.
2. Monthly: compare active referrals by sharing channel.
3. After campaigns: verify the exact referral URL used.
## Keep in mind
* No referrals is a valid state; the page starts empty until signups happen.
* Link accuracy matters more than volume when troubleshooting attribution.
* `Active` and `Inactive` status helps you understand referral quality, not just referral count.
## Where to go next
* [Profile General](/settings/profile/general)
* [Organization Billing](/settings/organization/billing)
* [Project Integrations](/settings/project/integrations)
# General
Source: https://docs.atomicagi.com/settings/profile/general
Manage your profile details, email status, password access, and appearance mode.
![Profile general settings showing full name, email verification, password access, and light/dark mode]()
Use this page for your personal account preferences. It does not change organization or project settings.
## Questions this page should answer
1. Is my profile identity correct in the app?
2. Is my account email verified?
3. Can I sign in with a password?
4. Is my preferred appearance mode set correctly?
## Before you update
* In the app, open your profile settings and select `General`.
* If your email is not verified, complete verification before relying on email-based flows.
* If you signed up with Google only, set a password before using email and password sign-in.
* Choose the display mode you want to use across the app.
## What this page controls
* `Full name` (read-only in this view).
* `Email` and verification badge (`Verified` or `Not verified`).
* `Password` access. Existing password users can open a dialog to change their password. Google-only users can open the same action to set their first password.
* `Appearance` mode (`Light` or `Dark`).
## What this page gives you
* A single place to confirm which Atomic user account is active.
* Email verification status for invite, notification, and account flows.
* Password setup/change access for non-Google sign-in.
* Appearance mode control for daily use.
## How to use this page
### Check identity and email status first
This page is the fastest place to confirm the account identity currently active in the workspace and whether the email is verified.
### Manage password access
Use the action on the `Password` row to open the password dialog. If your account already has a password, enter your current password and a new password to change it. If your account was created with Google only, enter and confirm a new password to enable email and password sign-in.
![Profile general settings with the password row and change action]()
![Password dialog opened from profile general settings]()
![Profile general settings after the password dialog closes successfully]()
### Set appearance mode for daily use
Switch between `Light mode` and `Dark mode` based on readability for your workflow.
### Treat this page as personal scope only
Nothing here changes team permissions, billing, or project-level configuration.
## Quick monthly checklist
1. Confirm full name and email match your active account.
2. Verify email status is green before invite and notification workflows.
3. Confirm you have password access if you need a non-Google sign-in option.
4. Re-check appearance mode if you changed devices or environments.
## What to fix first
| Pattern in Profile General | What it usually means | Recommended action |
| ---------------------------------- | ------------------------------------------ | ---------------------------------------------- |
| Email is not verified | Email-based flows may be unreliable | Complete verification before relying on emails |
| Google-only account needs password | Email/password login is not enabled | Set the first password from the password row |
| Wrong account is visible | Chrome/session is signed into another user | Switch accounts before changing project data |
| Appearance is hard to read | Mode does not match your working context | Switch light/dark mode |
## Keep in mind
* Profile settings apply to your user account only.
* Organization and project settings are configured in their own groups.
## Where to go next
* [MCP setup](/settings/profile/mcp)
* [Profile Affiliate](/settings/profile/affiliate)
* [Organization Members](/settings/organization/members)
* [Project Members](/settings/project/members)
# MCP setup
Source: https://docs.atomicagi.com/settings/profile/mcp
Connect Claude, ChatGPT, Codex, and other AI agents to Atomic through OAuth 2.1.
![Profile MCP settings showing the required MCP server URL with personal access tokens as an optional fallback]()
Connect supported AI agents through OAuth 2.1 to call Atomic tools with your existing workspace and project
permissions. Create a personal access token only for a client that cannot connect with OAuth.
MCP is available for projects on Starter and higher plans. A token remains personal, but Atomic checks the plan of
the project selected for each MCP request.
## Questions this page should answer
1. How do I connect my AI agent with OAuth?
2. Which MCP endpoint should my client use?
3. When do I need a personal access token?
4. How do I confirm tools are available in my client?
## Before you connect
* In Atomic, open `Profile settings` and select `MCP`.
* Use the MCP endpoint:
```txt theme={null}
https://app.atomicagi.com/api/mcp
```
Copy the MCP server URL from the page and paste it into your AI agent's custom connector setup. Clients with OAuth
support open Atomic sign-in and ask you to approve access. No personal access token is required.
## What this page gives you
* Your Atomic MCP endpoint.
* A copyable server URL used by OAuth-enabled custom connectors.
* Personal access token creation and revocation for clients without OAuth.
* The configuration pattern for other remote HTTP MCP clients.
Important: If you need a personal access token, treat it
like a password. Store it in your client config or secret manager, not in
shared docs or chat threads.
## How project access works
OAuth connections and personal access tokens inherit your user permissions. They cannot access projects you cannot
already open in Atomic, and they cannot use tools for projects below the Starter plan.
You do not need to put a project id in your client configuration. When a tool runs:
1. Atomic uses the only accessible project automatically when your account has access to one project.
2. If your account has access to multiple projects, Atomic returns a project chooser with each project id, name, domain, and workspace.
3. Ask the AI client to use the right project and rerun the tool with `projectId` in the tool arguments.
## Cursor
Cursor can connect directly to Atomic's HTTP MCP endpoint.
Open Cursor MCP settings and add this server to your `mcp.json`:
```json theme={null}
{
"mcpServers": {
"atomic-ai": {
"url": "https://app.atomicagi.com/api/mcp",
"headers": {
"Authorization": "Bearer atmcp_YOUR_TOKEN"
}
}
}
}
```
After saving, refresh the `atomic-ai` MCP server in Cursor. The tools list should load without any project-specific header.
## Claude custom connector
Claude and Claude Desktop can connect directly to Atomic using the remote custom connector flow. You do not need to paste an MCP token or configure OAuth client credentials.
1. In Claude, open `Settings` > `Connectors`.
2. Select `Add custom connector`.
3. Enter `Atomic AI` as the name.
4. Enter the remote MCP server URL:
```txt theme={null}
https://app.atomicagi.com/api/mcp
```
5. Leave the optional OAuth Client ID and OAuth Client Secret empty.
6. Select `Add`, then `Connect`.
7. Sign in to Atomic and approve the requested tool access.
![Atomic OAuth consent screen for connecting Claude to Atomic tools]()
Claude uses short-lived OAuth access tokens and refreshes them automatically. Disconnect the connector in Claude to revoke its OAuth grant.
## ChatGPT custom app
ChatGPT can connect to Atomic through an OAuth-enabled custom MCP app. You do not need to create or paste a personal MCP token.
1. Enable developer mode in ChatGPT.
2. Open `Settings` > `Apps` and create a custom app.
3. Enter the Atomic MCP endpoint:
```txt theme={null}
https://app.atomicagi.com/api/mcp
```
4. Choose OAuth authentication and let ChatGPT register its public client automatically.
5. Sign in to Atomic and approve the requested tool access.
Atomic accepts ChatGPT's current per-app callback URL and its documented legacy callback. The OAuth client remains restricted to the exact callback URI ChatGPT registered.
## Codex
Codex can add Atomic as a remote MCP server and authenticate through OAuth:
```bash theme={null}
codex mcp add atomic-ai --url https://app.atomicagi.com/api/mcp
codex mcp login atomic-ai
```
Open the authorization URL if prompted, sign in to Atomic, and approve access. Restart Codex after login so the
Atomic tools are available in a new session.
## Claude Desktop local configuration
For clients that use Claude Desktop's local `claude_desktop_config.json` rather than remote custom connectors, use `mcp-remote` to bridge to Atomic's HTTP MCP endpoint.
Add this server to `~/Library/Application Support/Claude/claude_desktop_config.json` on macOS:
```json theme={null}
{
"mcpServers": {
"atomic-ai": {
"command": "npx",
"args": [
"-y",
"mcp-remote@latest",
"https://app.atomicagi.com/api/mcp",
"--header",
"Authorization:${ATOMIC_MCP_AUTH_HEADER}"
],
"env": {
"ATOMIC_MCP_AUTH_HEADER": "Bearer atmcp_YOUR_TOKEN"
}
}
}
}
```
Fully quit and reopen Claude Desktop after editing the file.
## Claude Code
Claude Code can add Atomic as a remote HTTP MCP server from the terminal:
```bash theme={null}
claude mcp add --transport http atomic-ai https://app.atomicagi.com/api/mcp \
--header "Authorization: Bearer atmcp_YOUR_TOKEN"
```
Then run this inside Claude Code:
```txt theme={null}
/mcp
```
Confirm that `atomic-ai` is connected and that tools are available.
## Other MCP clients
Native and desktop MCP clients can use Atomic OAuth when they support dynamic client registration, authorization code flow with PKCE, and an HTTP loopback callback on `localhost`, `127.0.0.1`, or `::1`. Atomic validates and stores the exact registered callback URI.
Clients without OAuth support can use a personal bearer token:
Use Atomic as a remote HTTP MCP server when the client supports HTTP MCP:
```json theme={null}
{
"url": "https://app.atomicagi.com/api/mcp",
"headers": {
"Authorization": "Bearer atmcp_YOUR_TOKEN"
}
}
```
If the client only supports local stdio servers, use `mcp-remote` with the Claude Desktop pattern.
## Test the connection
You can test the token and endpoint with `curl`:
```bash theme={null}
curl -X POST https://app.atomicagi.com/api/mcp \
-H "Content-Type: application/json" \
-H "Authorization: Bearer atmcp_YOUR_TOKEN" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
```
A healthy response includes a `tools` array.
## Quick setup checklist
1. Choose your AI agent on the Atomic MCP page.
2. Add the Atomic endpoint in your client and complete OAuth sign-in.
3. Restart or refresh the MCP client.
4. Confirm the Atomic tools list appears.
5. Run a harmless read-only tool first.
6. If asked to choose a project, pass the returned `projectId` on the next call.
## Troubleshooting
### The client shows no tools
Check that the MCP token is active, the endpoint ends with `/api/mcp`, and the authorization header starts with `Bearer `.
### Claude Desktop says the config is invalid
Use the `command` and `args` format with `mcp-remote`. Claude Desktop does not accept the direct `url` and `headers` format used by Cursor.
### Claude custom connector does not open Atomic sign-in
Confirm that the connector URL is exactly `https://app.atomicagi.com/api/mcp` and that the optional OAuth Client ID and OAuth Client Secret fields are empty. Remove and add the connector again if Claude cached a failed setup attempt.
### The tool asks you to choose a project
Choose the project by name or domain. The AI client should call the tool again with the returned `projectId` in the tool arguments.
### A token was shared by mistake
Revoke it from `Profile settings` > `MCP` immediately and create a new token.
## Keep in mind
* MCP tokens inherit your user permissions.
* Claude custom connectors use OAuth and do not require a manually created MCP token.
* ChatGPT custom apps and standards-compliant native MCP clients can use OAuth without a personal MCP token.
* Tokens do not grant access to projects you cannot open in Atomic.
* Project selection is explicit when your account can access multiple projects.
* Revoke old tokens when you stop using a client.
## Where to go next
* [Profile General](/settings/profile/general)
* [Project Members](/settings/project/members)
* [Organization Members](/settings/organization/members)
# AI Agent Settings
Source: https://docs.atomicagi.com/settings/project/ai-agent
Manage project-level BYOK provider keys so agents can use the right model credentials for this project
![Project AI agent settings page with bring-your-own-key provider cards]()
Use this page to manage project-specific AI provider keys. It is the place for bring-your-own-key setups that should apply only to this project.
## Questions this page should answer
1. Which providers are configured for this project?
2. Which provider keys are enabled right now?
3. When should we use project keys instead of system credentials?
## Before you configure keys
* Confirm this project actually needs bring-your-own-key behavior.
* Decide who owns key rotation and billing for the provider account.
* Check whether the provider account supports the models your agents need.
* Keep the key available only long enough to paste it into Atomic.
## What this page gives you
* Provider cards for supported BYOK setups.
* Configure, enable, disable, test, edit, and delete actions.
* Automatic fallback behavior when project keys are not enabled.
## Providers currently supported
* `Anthropic`
* `OpenAI`
* `Google`
* `Qwen`
* `Moonshot`
## How to use this page
### Configure only the providers you actually need
Do not add every possible key by default. Add the providers required for the models or policies this project really uses.
### Test before enabling at scale
Save and test the key before relying on it in production workflows. A key can be configured but still wrong for the model or account you expect.
### Leave fallback behavior intentional
If a project needs strict provider isolation, confirm which keys are enabled and who owns rotation. If fallback is acceptable, document that decision internally.
## How to know setup worked
* The provider card shows the configured state you expect.
* The key test succeeds after saving.
* Agents that use the provider can complete a small test request.
* Errors mention prompt/model issues rather than authentication or provider access.
## Quick monthly checklist
1. Confirm each enabled key still has an owner.
2. Re-test keys after provider account or billing changes.
3. Disable providers this project no longer uses.
4. Rotate keys according to your security policy.
## What to fix first
| Pattern in AI Agent settings | What it usually means | Recommended action |
| ---------------------------------- | ------------------------------------ | ---------------------------------------------------- |
| Key saves but test fails | Provider credential or account issue | Re-check key value, account status, and model access |
| Agent errors mention auth/provider | Project key is invalid or disabled | Re-test and update the provider card |
| Multiple unused providers enabled | Configuration is broader than needed | Disable unused keys to reduce operational risk |
| Outputs changed after key rotation | Model/provider behavior may differ | Run a small comparison test before scaling usage |
## Keep in mind
* These keys are project-scoped, not workspace-global.
* Key rotation should be followed by a quick re-test.
* A configured key is not the same as an enabled key.
## Where to go next
* [Agents](/automation/agents/overview)
* [Teams](/automation/teams/overview)
* [Project Integrations](/settings/project/integrations)
* [Project Notifications](/settings/project/notifications)
# Brand kit
Source: https://docs.atomicagi.com/settings/project/brand-kit
Create structured brand guidance used by AI agents and workflows.
![Project brand kit settings with multiple brand kits, brand information, writing style, and writing sample fields]()
Use this page to keep brand context clear and consistent before you run AI tasks or publish content. A project can have more than one brand kit for different brands, products, campaigns, or writing contexts.
## Questions this page should answer
1. Which brand kit should agents use for this work?
2. Is brand information specific enough for high-quality outputs?
3. Do the writing style and sample match the current campaign?
## Before you use this page
* In the app, open Project settings and select `Brand kit`.
* Gather current positioning notes, audience guidance, and style rules.
* Confirm the target country for search analysis.
## What this page controls
* Brand kit name and default selection
* Website or domain
* `About the brand`
* `Brand mention variations`
* `Ideal customer profile`
* `Competitors`
* `Brand point of view`
* `Author persona`
* `Tone of voice`
* `User perspective`
* `Writing rules`
* `Writing sample`
* `Content featured image` style, palette, background wash, layout, typography, and logo defaults
* `Search location`
## What this page gives you
* Reusable brand and writing context for agents and workflows.
* A default brand kit for the current project.
* Domain-assisted generation when you want a researched starting point.
* Image-style defaults for content featured images.
* Search location context for market-specific analysis.
## Why this matters
* `Search location` controls market context in AI search tracking.
* Brand and audience fields are reused by AI agents and workflows.
* The default brand kit is used by current agent and workflow paths.
* Vague entries lead to generic recommendations and weaker outputs.
## How to use this page
### Choose or create a brand kit
Use the brand kit selector to switch between saved kits. Click `New kit` to create a manual kit. Mark a kit as default when it should be the main brand guidance for the project.
### Generate and save from a domain
Enter a website or domain and click `Generate and save`. Atomic AI opens the predefined Brand Kit agent, researches the site, and saves the researched values with the brand kit tool. Review the saved fields afterward and refine any assumptions that need a human decision.
### Write concrete guidance
Use specific statements, examples, and constraints. Avoid broad terms like "professional" without context.
### Separate identity from style
Use `Brand information` for who the brand is, who it serves, competitors, and point of view. Use `Brand writing style` for author persona, tone, perspective, and concrete writing rules.
### Add specific writing rules
Capture banned words, capitalization rules, phrases to avoid, punctuation preferences, and other constraints agents should follow.
### Link a writing sample
Add a writing sample title and URL when you have a published example that shows the desired style.
### Tune content featured images
Use `Content featured image` to set the style rules, color mode, palette, logo assets, text placement, and typography for generated blog cover images. Lower `Background wash` when the background color feels too strong over an inline generated visual.
### Keep audience and tone aligned
When your campaign audience changes, update both `Ideal customer profile` and `Tone of voice` together.
### Set search location intentionally
Choose the country that matches the market you are measuring. This keeps prompt tracking and analysis aligned.
### Save after strategic changes
Update this page whenever messaging or positioning changes so new outputs use current guidance.
## Quick monthly checklist
1. Confirm the correct brand kit is marked as default.
2. Refresh brand and audience fields after positioning updates.
3. Confirm search location still matches your active market.
4. Remove old phrasing that no longer fits your messaging.
## What to fix first
| Pattern in Brand kit | What it usually means | Recommended action |
| --------------------------------- | ------------------------------------ | ------------------------------------------------------ |
| Agent output is generic | Brand fields are too vague | Add concrete audience, POV, and writing rules |
| Output has wrong voice | Tone/sample is stale or incomplete | Update tone of voice and writing sample |
| AI search context feels off | Search location may not match market | Check country selection |
| Featured images feel inconsistent | Image defaults are underspecified | Tune style, palette, layout, logo, and background wash |
| Multiple kits conflict | Default kit may be wrong | Confirm the intended kit is marked default |
## Keep in mind
* This page defines qualitative context, not indexed page/document sources.
* Brand kits are project-scoped, not organization-global.
* Use Knowledge for long reference documents. Use Brand kit for concise brand and style guidance.
## Where to go next
* [Project Knowledge](/settings/project/knowledge)
* [Project Integrations](/settings/project/integrations)
* [Project Data Sources](/settings/project/data-sources)
# Data Sources
Source: https://docs.atomicagi.com/settings/project/data-sources
Connect GA4 and GSC for project analytics and search performance data.
![Project data sources page with Google Analytics and Search Console cards]()
Use this page to control the two core project data connections: GA4 and Google Search Console.
Important: Wrong GA4 or GSC mappings can look like traffic
loss. Verify the connected property and site before diagnosing performance.
## Questions this page should answer
1. Are GA4 and GSC connected for this project?
2. Which exact property/site is connected?
3. Are conversion events configured as expected?
4. Will connecting a different GSC site change the project domain?
## Before you use this page
* In the app, open Project settings and select `Data sources`.
* Ensure your Google account has access to the correct GA4 property and GSC site.
* Confirm the current project domain and property IDs before editing.
## What this page controls
* GA4 connection state and property metadata.
* GA4 conversion events list.
* GSC connection state and connected website URL.
* Edit/connect actions for both integrations.
* Explicit confirmation before a different GSC hostname replaces the project domain.
## What this page gives you
* A quick health check for the project’s analytics and search data.
* The exact connected GA4 property and GSC site.
* Conversion event visibility for attribution and reporting.
* Reconnect/edit paths when source selection is wrong.
## How to use this page
### Validate connection badges first
Each card shows `Connected` or `Not connected`. Resolve red states before analyzing any downstream reports.
### Audit the linked property/site identifiers
Check GA4 property name and GSC website URL to ensure data comes from the correct source.
### Review a GSC domain change before connecting
When you choose a GSC website, Atomic compares its hostname with the current project domain. Protocol, `www`, URL paths,
and port numbers do not create a mismatch.
If the hostnames differ, the action changes to `Connect and change domain`. Selecting it opens a confirmation dialog that
shows the current and selected domains.
![Change project domain confirmation shown while connecting Google Search Console]()
* Select `Cancel` to keep the current project domain and return to the GSC website picker without connecting the site.
* Select `Connect and change domain` to connect the GSC website and replace the project domain with it.
### Keep conversion events intentional
Use GA4 edit flows to keep conversion events aligned with your current funnel definitions.
## Quick weekly checklist
1. Confirm both cards are connected.
2. Verify GA4 property and GSC URL still match the project domain.
3. Review any proposed project-domain change before reconnecting GSC.
4. Review conversion event list for drift after tracking changes.
## What to fix first
| Pattern in Data Sources | What it usually means | Recommended action |
| ----------------------------- | ------------------------------------ | ------------------------------------------------ |
| GA4 is disconnected | Attribution and engagement data weak | Connect the correct GA4 property |
| GSC is disconnected | Search reports cannot be trusted | Connect the correct Search Console site |
| Property/site looks wrong | Data may belong to another project | Edit the connection before analyzing trends |
| Conversion events are missing | Reports may understate outcomes | Update GA4 conversion event selection |
| Recent tracking changed | Old assumptions may be stale | Re-check both cards and reports after the change |
## Team routine
1. Weekly: confirm connection health before reporting.
2. After tracking changes: re-check conversion events.
3. After domain or property changes: reconnect the affected source.
## Keep in mind
* Report quality depends on clean GA4/GSC mappings.
* Wrong property/site selection can look like traffic loss when it is actually mapping error.
* A different GSC hostname changes the project domain only after you confirm the change.
## Where to go next
* [Project Integrations](/settings/project/integrations)
* [Project Notifications](/settings/project/notifications)
# General
Source: https://docs.atomicagi.com/settings/project/general
Update project identity and perform project-level lifecycle actions.
![Project general settings with project name, domain, and delete project action]()
Use this page to keep the project name and primary website current, and handle destructive project actions intentionally.
Important: Project identity controls how teams interpret
every report. Confirm project and domain before editing nearby settings.
## Questions this page should answer
1. Do the project name and domain match the current website?
2. Do I have the right permission scope for project actions?
3. Is project deletion warranted or premature?
## Before you use this page
* In the app, open the project you want to edit, then select `General` in Project settings.
* Validate you are editing the intended project before saving identity changes.
* Confirm backup/export expectations before deletion.
## What this page controls
* Editable project name and primary domain for members with General settings write access.
* `Delete project` action for allowed roles.
## What this page gives you
* A project identity check before deeper setup work.
* A clear place for destructive project lifecycle action.
* A reminder that project-level setup belongs in the adjacent project tabs.
## How to use this page
### Update the project name or domain
Edit the project name or enter the full website URL, including `http://` or `https://`, then select `Save` beside the field. Each change is saved independently. Atomic stores the domain as an HTTPS host, without `www`, paths, query parameters, fragments, or ports.
Members without General settings write access can review these values but cannot edit them.
### Use delete only for true decommissioning
Deleting a project is destructive; use it only when the project is intentionally retired.
### Route operational setup to other tabs
Data sources, members, and publishing are configured in their dedicated project tabs.
## Quick monthly checklist
1. Confirm the domain and project name still match live operations.
2. Revalidate who can run destructive actions.
3. Move routine configuration work to the correct project tabs.
## What to fix first
| Pattern in Project General | What it usually means | Recommended action |
| ------------------------------- | -------------------------------- | ------------------------------------------------- |
| Domain/name does not match work | You may be in the wrong project | Switch project before editing settings |
| Delete is being considered | Project may need archival review | Confirm backups, ownership, and downstream impact |
| User is looking for analytics | Wrong settings tab | Open Data Sources or Reports instead |
| User is looking for access | Wrong settings tab | Open Project Members |
## Team routine
1. Monthly: confirm project/domain identity.
2. Before major audits: verify the correct project is selected.
3. Before deletion: confirm downstream data, members, workflows, and reports.
## Keep in mind
* Changing the project domain does not reconnect Google Search Console, analytics, or report publishing domains. Review those connections separately when a website moves.
* General is identity/lifecycle, not integrations or analytics config.
## Where to go next
* [Project Members](/settings/project/members)
* [Project Data Sources](/settings/project/data-sources)
* [Project Integrations](/settings/project/integrations)
# Integrations
Source: https://docs.atomicagi.com/settings/project/integrations
Manage project Slack and publishing connections so delivery targets and notifications stay correctly wired
![Project integrations page with Slack and publishing connections]()
Use this page to manage project-level external connections. It combines Slack bot setup with publishing connections such as WordPress and Webflow.
Important: Verify connection state after every credential
or channel change. A saved integration is not always a working integration.
## Questions this page should answer
1. Is Slack connected to the correct channel for this project?
2. Which publishing connections are active and verified?
3. Which integration needs reconnect, edit, verify, or removal?
## Before you connect integrations
* Confirm you are editing the intended project.
* Gather credentials or admin access for each destination.
* Decide which Slack channel should receive project activity.
* Verify publishing destinations in a test-safe way before relying on them.
## What this page gives you
* Slack connection state and channel selection.
* Reconnect guidance when the Slack app is missing required mention scope.
* Publishing connection list and verification state.
* Add, edit, delete, and verify flows for publishing targets.
## Slack section
Use the Slack section to connect the project bot, select its channel, and confirm the channel used for project mentions.
If the page shows `Reconnect required`, remove the old connection and connect Slack again so the project has the required mention scope.
## Publishing connections section
Use publishing connections for delivery targets such as WordPress and Webflow.
* Add a connection only after credentials are ready.
* Verify each connection before live publishing.
* Remove stale destinations to prevent accidental publishing mistakes.
## How to know setup worked
* Slack shows the expected channel and no reconnect warning.
* Publishing destinations show verified connection state.
* A test publishing or verification action succeeds before production use.
* Notifications and workflow delivery reference the intended project channel or destination.
## Quick monthly checklist
1. Confirm Slack still points to the active team channel.
2. Verify publishing connections after password, API key, or CMS changes.
3. Remove destinations tied to old campaigns or retired sites.
4. Check notification settings after integration changes.
## What to fix first
| Pattern in Integrations | What it usually means | Recommended action |
| ------------------------------- | ----------------------------------------- | ---------------------------------------------------- |
| Slack reconnect is required | Required Slack scope or install changed | Reconnect Slack for this project |
| Publishing verification fails | Credentials or destination config drifted | Update credentials and verify again |
| Wrong channel receives messages | Project channel selection is stale | Change the Slack channel and test notifications |
| Many stale destinations exist | Publishing risk is higher | Remove inactive connections |
| Workflow delivery fails | Integration setup may be incomplete | Verify the destination before debugging the workflow |
## Team routine
1. Monthly: verify Slack and publishing destinations.
2. After credential changes: re-test the affected connection.
3. Before publishing workflows: confirm destination and project match.
## Keep in mind
* Slack is project-scoped here, even if the entry point begins from organization permissions.
* A connected integration is not always a verified integration.
* Credential changes should always be followed by a re-check.
## Where to go next
* [Project Data Sources](/settings/project/data-sources)
* [Project Notifications](/settings/project/notifications)
* [Brand kit](/settings/project/brand-kit)
# Knowledge
Source: https://docs.atomicagi.com/settings/project/knowledge
Manage OKF knowledge bases, folders, and markdown concepts for a project.
![Project Knowledge page with knowledge bases, folders, concepts, and markdown detail panel]()
Use Knowledge to keep project context readable for people and agents. Each project can have one or more knowledge bases. Each knowledge base contains folders and markdown concepts.
## Questions this page should answer
1. Which knowledge bases exist for this project?
2. Which folders and markdown files are available and ready for agents?
3. Which files and exact published passages support a search?
## Before you use this page
* Open the project and select `Knowledge`.
* Decide whether the content belongs in an existing knowledge base or a new one.
* Keep each concept focused on one page, document, story, note, or generated output.
## What this page gives you
* Knowledge-base cards with purpose, file count, readiness, warnings, and the last successful index.
* One expandable file tree that keeps the knowledge-base root visible.
* Row-level summaries, tags, processing state, and recovery actions.
* Two-stage Test Search scoped to the selected knowledge base.
* File preview with published-version metadata and exact, server-resolved citations.
## Knowledge bases
A knowledge base is the top-level bundle. Use separate knowledge bases when the content has different ownership or purpose, such as product docs, customer stories, pricing notes, or implementation references.
Every knowledge base needs a purpose description. Atomic shows it on the knowledge-base card and uses it when interpreting searches and preparing file metadata. The description guides retrieval inside that knowledge base; it never grants access or causes Atomic to search a different knowledge base.
When you create or edit a knowledge base, write a short purpose that answers:
* What belongs here?
* What should be excluded?
* How should agents use this material?
For example: `Approved customer outcomes and proof points for sales and marketing content. Exclude internal roadmap notes and unverified claims.`
Changing the purpose marks its concepts as stale so their compact metadata can be refreshed. It does not rewrite file content or create duplicate concepts.
## Share and restore a Knowledge view
Knowledge navigation is addressable. Opening a knowledge base changes the path to `/knowledge/[knowledgeBaseId]`. The URL also keeps the selected folder and search query:
```text theme={null}
/knowledge/42?folder=17&q=pricing
```
You can copy that URL, refresh it, or open it in a new tab to restore the same state. The recipient must already have access to the project and selected knowledge base. Invalid, deleted, cross-project, or inaccessible IDs show an authorization-safe unavailable state and never broaden search scope.
## Browse folders and files
Folders organize concepts inside a knowledge base. A folder can contain subfolders and concepts. Each folder has an `index.md` view generated from its children.
![Expandable Knowledge file tree with nested folders, markdown files, summaries, and readiness]()
Open a folder row to reveal its direct children in the same table. Atomic loads each branch when you first expand it and keeps other open branches visible. Use the disclosure control or the Left and Right arrow keys to collapse and expand a focused folder.
The columns keep the meaning of each row visible:
* `Name` shows the human-readable name and canonical path.
* `Summary` shows generated context, classification, and tags for files.
* `Status` distinguishes folders from pending, processing, ready, stale, and failed files.
* `Added` shows when the item entered the knowledge base.
Opening and closing folders never changes where new data will be created.
## Concepts
A concept is a markdown file with frontmatter. The markdown file is the source of truth. The app stores a database id for safe moves and renames, but the OKF path stays visible for import and export.
## Indexing status
Atomic processes every new or changed file before agents can retrieve it.
![Knowledge file table showing an indexed markdown file and its row actions]()
* `Added for processing` means the file is saved and waiting for the knowledge worker.
* `Syncing` means Atomic is normalizing, enriching, segmenting, or indexing the current version.
* `Ready` means every required step completed and the current version is indexed for agents.
* `Stale` means the file or its knowledge-base context changed after the last successful index.
* `Failed` means one processing step failed.
While processing runs, the file row shows the current step and progress. Open the row menu and select `Retry indexing` after a failure. Select `Re-index` to rebuild an otherwise healthy file.
Atomic publishes a file version only after every required step finishes. Search does not use partially processed versions.
## Add data
Open a knowledge base and select `Add data`. Choose one of the currently available actions:
* `Create folder` adds an organizational folder.
* `Create file` opens the markdown Write/Preview editor.
Both forms show a `Location` field. The knowledge-base root is the default, even when folders are expanded. Choose a folder explicitly when the new item belongs in a nested location.
![Paste Text dialog showing a markdown file in Preview mode]()
1. Add an optional file name. If you leave it empty, Atomic uses `Pasted Text`.
2. Write or paste markdown in the `Write` tab.
3. Select `Preview` to check headings, lists, links, and other formatting.
4. Select `Add for processing` to save the note at the displayed location.
5. Wait for the status to move from `Pending` or `Processing` to `Ready`.
The other file and connected-source imports are not available from this flow yet.
## Search
Use `Search all knowledge` on the Knowledge overview to search every Knowledge base available in the current project.
The query stays in the URL, and results are grouped by Knowledge base with the matching file path and exact published
supporting passages. Opening a file or citation navigates into its Knowledge base. Clear the query to return to the
normal Knowledge overview.
Use `Search files` beside the knowledge-base title and press Enter to test what agents can retrieve. Clear the field to return to the file table.
![Knowledge search showing candidate files, supporting passages, and a resolved published citation]()
Knowledge Search runs in two stages:
1. `Candidate files` checks compact published metadata such as titles, descriptions, paths, tags, summaries, classifications, headings, and the knowledge-base purpose. It does not read every complete file.
2. `Supporting passages` searches only the published segments from those candidate files. It shows the exact text and location that an agent can cite.
Results remain inside the selected knowledge base and optional folder. Search uses only complete indexed versions. When a file is stale or processing, the result says which last complete version is being served.
Select `Open citation` to open the published file version and highlight the supporting segment. Markdown citations show the heading and line range. Page, table, spreadsheet, and OCR locators appear when the source extractor provides them. If Atomic cannot render the original source, the preview falls back to the canonical extracted passage and shows its locator metadata.
Citation links contain an opaque stored segment identity. Atomic resolves that identity on the server and does not trust a model-generated URL, page number, or range as citation truth.
## Manage knowledge with MCP
Atomic exposes the same Knowledge base commands and permission checks through MCP. IDs are always explicit. A tool never changes an empty scope into every knowledge base in the project.
Start with these read tools:
| Tool | Use it for |
| ------------------------------- | --------------------------------------------------------------------- |
| `list_knowledge_bases` | Find allowed base IDs, purposes, counts, readiness, and last indexing |
| `get_knowledge_base` | Inspect one base and its root folder/file summary |
| `list_knowledge_folder` | Browse one base root or folder with cursor-based pagination |
| `get_knowledge_file` | Read canonical markdown, metadata, readiness, and published segments |
| `search_knowledge` | Search explicit base IDs and return exact citation locators |
| `get_knowledge_indexing_status` | Check the current processing step, progress, readiness, or error |
Use these write tools to manage canonical knowledge:
| Tool | Use it for |
| ------------------------- | ------------------------------------------------------------------------ |
| `create_knowledge_base` | Create a base with a required purpose description |
| `update_knowledge_base` | Rename a base or change its purpose and re-index affected context |
| `create_knowledge_folder` | Add a root or nested folder |
| `update_knowledge_folder` | Rename, describe, or move a folder |
| `add_knowledge_file` | Add canonical markdown to an explicit base and optional folder |
| `update_knowledge_file` | Edit or move canonical content and start the required indexing lifecycle |
| `reindex_knowledge_file` | Re-index one file with a caller-stable idempotency key |
The delete tools are `delete_knowledge_file`, `delete_knowledge_folder`, and `delete_knowledge_base`. Each requires `confirmation: "DELETE"`. Base deletion cascades to its folders, files, processing history, and derived indexes. Folder deletion works only after its contents are moved or deleted.
To search, pass at least one ID:
```json theme={null}
{
"query": "What does the Starter plan include?",
"knowledgeBaseIds": [12],
"limit": 8
}
```
`search_knowledge` first selects candidate files from compact published metadata. It then returns exact published segments with citation IDs, paths, headings, and line ranges. `Indexed` means the current canonical version completed every required publication step. `Stale` can continue serving its last complete version while the replacement is processing.
## Check content against project knowledge
Use the `check_knowledge_base_alignment` MCP tool before publishing content that makes claims about your company, product, services, or policies. Pass the draft in `content` and an explicit `knowledgeBaseIds` allowlist. You can also set `maxClaims` from 1 to 50; the default is 20.
The tool separates the draft into factual claims and checks each claim against matching excerpts from the selected indexed knowledge bases. It returns one of these verdicts for every claim:
* `supported` means an indexed source directly supports the claim.
* `contradicted` means an indexed source directly conflicts with the claim.
* `not_found` means the available sources do not provide enough evidence. Missing evidence does not count as a contradiction.
Each supported or contradicted verdict includes the source title, URL when available, matching excerpt, and retrieval score. Review contradicted claims first. Then decide whether `not_found` claims need a draft correction or a new knowledge source.
## What to fix first
| Pattern in Knowledge | What it usually means | Recommended action |
| --------------------------- | ----------------------------------- | --------------------------------------------- |
| Important pages are missing | SEO audit or sync coverage is stale | Run SEO audit, then sync new pages |
| Documents are outdated | AI may cite old guidance | Delete stale docs and upload current versions |
| `Never` appears repeatedly | Sources may be pending or failed | Re-run sync or check file/source validity |
| Outputs ignore key context | Source coverage is incomplete | Add the missing page or document |
| Too many old sources exist | Retrieval quality may degrade | Remove obsolete files and pages |
## Quick weekly checklist
1. Add new pages, docs, and notes as focused concepts.
2. Keep folder names clear enough to scan.
3. Search for priority terms and confirm the expected concepts appear.
4. Remove stale concepts after moving or replacing them.
## Keep in mind
* Project permissions still control who can view or edit knowledge.
* Agent and MCP operations require explicit knowledge-base IDs and enforce the same project role as the UI.
* The concept markdown is canonical. Avoid storing important context only in source metadata.
## Where to go next
* [Brand kit](/settings/project/brand-kit)
* [Interlinking](/data/technical/interlinking)
* [Agents](/automation/agents/overview)
# Members
Source: https://docs.atomicagi.com/settings/project/members
Manage project-level access, role visibility, and permission granularity.
![Project members table with invite action and member status badges]()
Use this page to manage who can work inside this specific project.
Important: Project access should match current work. Remove
stale access before it becomes a security or ownership problem.
## Questions this page should answer
1. Who currently has access to this project?
2. Which members are pending vs accepted?
3. How should I scope permissions per person?
## Before you use this page
* In the app, open Project settings and select `Members`.
* Check member seat capacity if invites are blocked.
* Confirm whether changes should be project-only or organization-wide.
## What this page controls
* Project member list (`Email`, `Role`, `Status`, `Actions`).
* Invite flow for adding members to this project.
* Permission editing dialog for project-scoped feature/settings access.
* Remove member from project action.
## How to use this page
### Invite with scope in mind
Invite only users who need this project, then verify status changes from `Pending` to `Accepted`.
For the full invite and permission model, use the detailed guide in
[Organization Members](/settings/organization/members#invite-members-flow) (role behavior, project assignment, and `None` / `View only` /
`Full` scope definitions for each permission).
### Edit permissions at project scope
Use the edit permissions action to control this project’s feature/settings access without changing all organization access.
### Keep the list operationally clean
Remove members who no longer need project access to reduce risk and noise.
## Quick weekly checklist
1. Clear stale pending invites.
2. Review elevated project permissions.
3. Remove contributors no longer active on this project.
## What to fix first
| Pattern in Project Members | What it usually means | Recommended action |
| --------------------------- | --------------------------------------- | ------------------------------------------ |
| Invite remains pending | User has not accepted or email is wrong | Resend or recreate the invite |
| Member has broad access | Permissions may exceed project need | Narrow feature/settings permissions |
| User needs all projects | Project-level invite may be too narrow | Review Organization Members |
| Seat capacity blocks invite | Billing/seat limits need attention | Check organization billing before inviting |
| Former contributor remains | Access cleanup is overdue | Remove project access |
## Team routine
1. Weekly: clear pending invites and remove stale contributors.
2. Monthly: review elevated permissions.
3. After role changes: confirm project-level scope still matches the user’s work.
## Keep in mind
* Project membership is narrower than organization membership.
* Seat limits can still impact invite operations.
## Where to go next
* [Organization Members](/settings/organization/members)
* [Project General](/settings/project/general)
* [Project Integrations](/settings/project/integrations)
# Notifications
Source: https://docs.atomicagi.com/settings/project/notifications
Schedule recurring reports, digests, and SEO audit notifications so the right people get the right cadence
![Project notifications page with recurring report and audit schedules]()
Use this page to manage recurring project notifications. This is where you schedule performance reports, weekly digests, and SEO audit notifications.
## Questions this page should answer
1. Which recurring notifications are active?
2. Are cadence, timezone, and next run correct?
3. Are email or Slack alerts enabled only where they add value?
## Before you schedule notifications
* Decide who will act on each notification.
* Confirm the project timezone and reporting cadence.
* Check [Project Integrations](/settings/project/integrations) if Slack notifications are expected.
* Avoid scheduling reports before data sources are connected and stable.
## What this page gives you
* Separate schedule controls for `Performance report`, `Weekly digest`, and `SEO audit`.
* Frequency, day, monthly option, time range, and timezone settings.
* Notification channel toggles where supported.
* Visibility into the next run and the current schedule state.
## How to read notification schedules
Review each schedule as a commitment to send attention somewhere.
* `Frequency`: how often the report or audit should run.
* `Day` and `time`: when the team expects to receive it.
* `Timezone`: which team/location the run is aligned to.
* `Channel toggles`: where completion or report notifications are delivered.
* `Next run`: the fastest way to verify the saved schedule.
Use this rule:
* Weekly is the default for most operating reports.
* Monthly is better for leadership rollups.
* Daily should be reserved for issues that someone actively reviews every day.
## How to use this page
### Match cadence to decision rhythm
Weekly or monthly schedules are enough for most reporting. Higher-frequency schedules should only exist when a team is actually reacting to them.
### Review timezone and next run together
Timezone drift is a common reason reports feel broken even when the schedule technically saved correctly.
### Limit notification noise
Enable notifications only for outputs that someone will read and act on. More alerts usually reduce attention instead of increasing it.
## How to know setup worked
* The schedule shows the intended next run.
* Timezone matches the receiving team.
* Notification channel toggles match the expected delivery path.
* Slack-related options are backed by a connected project Slack integration.
## Quick weekly checklist
1. Check active schedules for next-run accuracy.
2. Pause schedules with no clear reader or owner.
3. Confirm Slack/email delivery still matches team workflow.
4. Review SEO audit notification cadence after major site changes.
## What to fix first
| Pattern in Notifications | What it usually means | Recommended action |
| --------------------------- | ------------------------------------- | --------------------------------------------- |
| Reports arrive at odd times | Timezone or day is wrong | Update schedule timing and verify next run |
| Team ignores notifications | Cadence or channel is too noisy | Reduce frequency or route to a better channel |
| Slack alerts do not arrive | Project Slack integration may be weak | Check Project Integrations first |
| SEO audit runs too often | Monitoring cadence exceeds decisions | Move to weekly or monthly |
| Old schedules still run | Ownership changed | Pause or delete stale schedules |
## Team routine
1. Weekly: review schedules that triggered action.
2. Monthly: remove alerts that no one used.
3. After team changes: re-check timezone, channel, and owner.
## Keep in mind
* This page is for recurring reports and alerting, not for full agent-conversation automations.
* Slack notification behavior depends on the project Slack connection.
* A schedule should follow operating cadence, not tool availability.
## Where to go next
* [Project Integrations](/settings/project/integrations)
* [Automations](/automation/automations/overview)
* [Workflows](/automation/workflows/overview)
* [Technical overview](/data/technical/overview)
# Product images
Source: https://docs.atomicagi.com/settings/project/product-images
Upload and describe first-party visuals that content workflows can reuse.
![Project Product Images settings with upload metadata fields, search, and saved product images]()
Use Product Images to keep approved screenshots, dashboard examples, brand visuals, customer logos, and feature images ready for content workflows. The library belongs to the current project.
## Before you add an image
* Confirm your team owns the image or is allowed to reuse it.
* Only upload assets approved for public use. Atomic stores each library image at a public, non-expiring URL so published content can display it.
* Choose a clear image that supports a specific product, feature, or use case.
* Prepare a concrete title and description. These two fields are required.
## Add an image
1. In Project settings, open `Product images`.
2. Click `Choose image` and select an image file.
3. Add a title and description.
4. Add optional tags, use cases, product or feature, and source or owner details.
5. Click `Add image`.
Use metadata that describes what is visible and where the image is useful. For example, use `Attribution dashboard` as the title, describe the metrics shown, add `analytics, attribution` as tags, and add `blog screenshots` as a use case.
## Find and update images
Use `Search metadata` to filter the library by title, description, tags, use cases, product or feature, and source or owner. Edit an item when its description or intended use changes.
To replace the image file, delete the current item and upload the replacement. Deleting an item permanently removes it from the project library and Atomic storage.
## How workflows use the library
`Enrich content images` searches the active project's Product Images before looking for external visuals. When a saved image is a strong match for a section, the workflow inserts its existing Atomic storage URL into a new active content version.
Product Images are first-party assets, so they are not imported again and do not need a public source-credit line. If no saved image is relevant, the workflow can still find an external image, import it into Atomic storage, and add a visible source credit.
## Permissions
Product Images uses the project's Brand kit permission. Members with read access can view and search the library. Members with write access can upload, edit, and delete images.
These permissions control library management and retrieval inside Atomic. They do not restrict access to an uploaded image's public URL. Anyone who has that URL can load the image, so do not upload confidential screenshots, private customer data, or assets that are not approved for public publishing.
## Keep in mind
* The library is project-scoped. Images from another project are not returned.
* Specific metadata makes retrieval more reliable than broad labels such as `screenshot`.
* The workflow skips weak matches instead of forcing a product image into an unrelated section.
## Where to go next
* [Workflows](/automation/workflows/overview)
* [Brand kit](/settings/project/brand-kit)
* [Knowledge](/settings/project/knowledge)