How it works
The weekly agent loop, the schedule, ticket statuses, and where a human steps in.
AgenticSEO runs a loop for each site. Research and the AI visibility check find work, the writer turns work into drafts, the QA gate decides what may go live, and audits find pages to improve. Everything moves through tickets on the Ticket board, so you can see what each agent did and why.
The loop
- Research finds keyword opportunities and writes a brief per ticket. See Keyword research.
- AI visibility lists buyer prompts, checks ChatGPT and Gemini, and turns gaps into tickets. See Prompts.
- Writer writes an article from the brief. It does not publish. See Writer.
- QA gate runs code checks and an AI editor review. A fail goes back to the writer. See QA gate.
- Publish sends only QA-passed drafts to your CMS. See Publishing.
- Audits compare the last 28 days with the 28 days before and make improvement tickets. See Audits.
- Conversions from PostHog or Google Analytics 4 rank converting pages and topics first. See Analytics.
- Lessons you add go into every agent prompt. See Lessons.
Two jobs run beside the loop: the daily Search Console pull for the Search performance page, and X distribution after each article is published. See X.
The schedule
Set the schedule in Settings > Weekly schedule.
| Setting | What it does | Default |
|---|---|---|
| Run the agents on a schedule | Turns scheduled research, audits and writer pickup on or off | On |
| Keyword research day (step 2) | Day research runs | Monday |
| Audit day (step 6) | Day the audit runs | Thursday |
| Hour (UTC) | Hour both jobs start | 6 |
| New tickets per research run | Most new tickets one research run creates | 5 |
| Max tickets per audit | Most tickets one audit creates | 5 |
| Articles the writer picks up per week | Weekly writer limit | 3 |
| Writer picks up the next to-do ticket by itself (at most one a day) | Automatic writing | On |
The AI visibility check runs weekly, right after keyword research, when Run the AI prompts check every week is on. The Search Console daily pull runs once a day even when the schedule is off.
Tip: You can run any job by hand. Use Run keyword research, Run audit or Sync conversions on the Overview page, Run now in the AI prompts section of Insights, or Write now on a ticket.
Ticket statuses
Every piece of work is a ticket. The board groups statuses into columns.
| Board column | Statuses | Meaning |
|---|---|---|
| To do | To do, Backlog | Waiting for the writer |
| In progress | Writing, In QA, Rewriting, Publishing | Writer and QA at work |
| Ready | Ready | Passed QA, not live yet |
| Needs human | Needs human | QA kept failing, or technical work |
| Published | Published | Live (last 30 days on the board) |
The normal path is: To do, Writing, In QA, (Rewriting, In QA) as needed, Ready, Publishing, Published.
A ticket moves to Needs human when QA fails more times than Max rewrites before a human looks allows (default 3). Technical findings from audits start in Needs human, because the writer does not handle them.
Ticket types
| Type | What the writer does |
|---|---|
| New article | Writes a new page |
| Refresh | Updates an existing page in place |
| Landing page | Writes a landing page |
| Comparison | Writes a comparison page |
| Internal links | Adds internal links to an existing page |
| Title/meta | Improves the title and meta description |
| Technical | None. A person fixes it, then clicks Mark done |
Where a human steps in
You choose how much control you keep:
- Review first: turn off Publish automatically after QA passes. Ready drafts wait for your Publish click.
- Review links: share a draft with a reviewer who has no account. See Review links.
- Lessons: teach the agents once, and every later prompt follows it.
- Stop: stop a running agent from the ticket page or Agent activity. It stops after its current step. Work saved before that is kept.
Note: An article is only published after it passes QA. The Publish button refuses a draft whose latest QA report failed.
See what the agents did
- Agent activity shows every run with its full log and AI usage for that run.
- Insights > Agent reports shows summaries of research, audits and AI prompt checks, including rejected keywords.
- AI usage shows tokens and searches per job, ticket and day. See AI usage.
Did this page help? Tell us at support@agenticseo.com.