Platform · Content Routines
One article at a time, about the prompt you are losing.
A content routine is a standing instruction. Daily or weekly, it finds the tracked prompt where your brand is weakest and reads what the AI engines answer and cite for it. What it produces is one draft, and that draft waits for a person to review it: nothing publishes itself.
What a routine is
A standing instruction, not a content calendar
Most content plans start from a keyword list and a guess. A routine starts from your tracking. You set it once: how often it runs, which of three measures counts as weakest, and which prompts it may choose from.
On schedule it picks the single tracked prompt where your brand is losing most, then researches it and writes one draft meant to become the page the engines cite for that question. The draft lands in a review queue, where a person approves it or rejects it with a reason. The next run starts over with whichever prompt is weakest then.
Selection
How a routine picks its prompt
Every run scores the prompts its filters allow with the same numbers your reports use, then takes the strongest candidate for the strategy you chose.
Three strategies, one rule each
Lowest visibility takes the prompt with the lowest mention rate, so it always has a candidate. Competitor gap looks for the prompt where competitors are mentioned more often than you are, and skips the run when you are ahead everywhere. Mentioned late is for the prompts you do appear in, but near the end of the answer.
A pool with enough evidence
A prompt qualifies only with at least five AI answers in the window, so the pick rests on a pattern rather than a single response. Narrow the pool by window, tag, or persona, or point the routine at a saved report and it borrows those filters live.
A skipped run is normal
The routine leaves alone any prompt it already covered inside the window, until that window rolls over. Rejecting an article frees its prompt for later. When nothing qualifies the run skips, and the history says why, such as "No prompt had competitors ahead of your brand."
A run, start to finish
Research first, then one draft
The writer never starts from a blank page. First the run gathers what it needs from your own tracking: the last three AI answers for that prompt, the ten pages and domains they cite most, the competitors they name and how often, the prompt’s persona, your brand description and owned sites, and the brand voice and standing instructions you set on the routine.
Between the pick and the draft, four agents work in order, each with one job.
- Researcher: searches the web, reads your own site and the pages already cited, and verifies every source before returning the gap, the angle, and an outline.
- Writer: answers the question in the first paragraph, leaves out outbound links and competitors entirely, and names your brand once or twice where it fits, though never in the opening.
- Reviser: checks the draft against the brief and applies our rules against machine-sounding prose.
- FAQ writer: adds four to six questions the body has no heading for.
Review
A person decides, every time
Every draft waits for a reviewer. Edit the text by hand, or click Revise with an instruction and the editing agent rewrites the document. Revert last change undoes one step. The sources list is there for you rather than the reader: the article itself never links out.
An article is Draft, Approved, or Rejected, and the last two are final. Approving locks the text for good, and rejecting asks for a reason and frees the prompt for a later run. Neither one publishes anything: approval only marks an article ready to publish, and in Managed GEO our team puts it on your site as part of the engagement.
Why it matters
Content aimed at the question you are losing
The engines already show which answers leave you out and which sources they trust instead. A routine writes from that record, on a schedule, and a person signs off before anything reaches your site. Every article exists for a reason you can point to in your report.
1
Article per run
Always the strongest candidate for the strategy, and at most one scheduled run a day.
Daily
Or weekly, your choice
The platform picks the run time, so routines spread across the week instead of all firing at once.
3
Ways to pick a prompt
Lowest visibility, Competitor gap, or Mentioned late.
5
AI answers before a prompt qualifies
Counted inside the routine’s window. A new workspace produces nothing until the tracking has run.
Inside our GEO work
Where routines sit in the method
Routines are the writing arm of Optimize. They read what Discover collects and give Detect something new to measure.
01 · DISCOVER
Daily tracking supplies what a routine needs: the scorecard that ranks your prompts, the answers themselves, and the pages they cite. Without it there is no pool to pick from.
02 · OPTIMIZEThis feature
Each run turns your weakest prompt into one reviewed draft, written from what the engines say and cite, and the approved articles become the content we ship.
03 · DETECT
Once an article is live, the same prompt keeps running every day, so the report shows whether its mention rate or position moved, and the routine moves on to the next weakest prompt.
Questions about content routines
What clients ask about the drafts
Does a content routine publish articles automatically?
No. A routine writes one draft per run and stops there. A person approves or rejects every article, and approval only locks the text and marks it ready to publish. The platform has no CMS connection. In Managed GEO our team publishes approved articles on your site as part of the engagement.
Who approves an article?
Whoever holds the review queue. In Managed GEO that is our strategists, who edit, approve, and publish on your behalf. In GEO Consulting & Trends your own content team approves, working from the brand voice and standing instructions we set on each routine. Every approval and rejection is recorded with the name of the person who made it.
How does a routine choose what to write about?
From your tracking rather than a keyword list. Each run scores the prompts its filters allow, keeps those with at least five AI answers in the window, drops any this routine already covered, and applies the strategy you chose: lowest mention rate, largest competitor gap, or lowest position score. Ties go to the prompt with more competitor mentions, then the one with more search volume.
Will the articles sound like us?
The writer works from a brand voice you describe in your own words on the routine, plus standing instructions such as claims to avoid, a call to action, or sections you always want. If you already publish articles, the platform can draft that voice from their URLs. A reviser then checks the draft against the brief and our rules against machine-sounding prose, and you can still edit by hand or send it back with an instruction.
How many articles will we get a month?
At most one per run, so a weekly routine yields up to four or five a month and a daily one up to thirty. Runs skip when no prompt qualifies, which is normal, so the real count depends on how many tracked prompts sit inside the routine’s window with enough answers. Accounts start with three routines, and we size that to the engagement.
Can we reject an article, and what happens then?
Yes. Reject asks for a reason, which stays on the article, and the decision is final: a rejected article cannot be edited or approved later. That releases the prompt it was written for, so a later run can pick it up again and start over with fresh research.
Related features
What a routine reads from, and who it is for
- The daily runs and the scorecard every routine ranks its prompts with.
Prompt Tracking
- The competitors behind the Competitor gap strategy, and the context the writer is told to leave out of the article.
Competitor Research
- A persona narrows the pool and tells the researcher who is asking and who will read.
Visibility by Persona
- No marketing team: we run the audit, write and publish the content, and track what AI says about you, on one monthly retainer.
Founders
See what AI is already saying about your brand.
A free audit shows you exactly where you stand across ChatGPT, Gemini, Perplexity, Copilot, Claude, Grok, and AI Overviews, and what it would take to close the gap.