How do AI engines decide what to cite?
By assembling an answer from sources whose claims they can corroborate against each other, preferring pages that answer the question directly in a self-contained chunk, name the entities involved explicitly, and carry claims that appear elsewhere too. The engines' own published guidance and observed behaviour agree that quality and verifiability, not production method, are the variables. A page that answers a real question first, with claims that can be checked, is more likely to be cited than a longer page that buries the answer or states figures nobody else states.
The piece on structuring content for LLM citation and the entity SEO piece give the positive standard.
What does a Search Visibility FTE actually do?
Every week it samples an approved prompt set, the questions your buyers ask, across the search and AI engines you selected, recording the date and whether your site was cited. From the sample and the connected console it proposes topics with a worth-it verdict: is the question asked, do you already answer it, is a competitor cited, can you answer it from approved sources. Approved topics are drafted answer-first with entities named, every claim through the fact gate, and published on your approval through your own route, as a pull request for a site built as code or through the endpoint package or a CMS connection. Then it measures again and reports per engine as a range.
The how-it-works piece and the ninety-day playbook cover the loop; the day-with story shows a day of it.
How is visibility measured honestly?
As citation presence: the share of the fixed prompt set for which your site is cited in an engine's answer, sampled weekly per engine with dates, reported per engine as a range over the cycle because the same question gets different answers on different days. Console impressions and clicks for the target queries sit beside it where the console is connected. Reach on platforms the FTE cannot sample, attribution of enquiries or sales to a page, and reader satisfaction are marked unmeasured rather than estimated, which is what makes the measured numbers credible. Movement is read across cycles, and a conclusion is drawn at the quarterly review.
The metric explainer on citation presence gives the computation with the inputs blank.
Are AI-written pages penalised?
Published guidance from major engines targets unhelpful, thin, and unoriginal content regardless of how it was produced, and AI engines cite claims they can corroborate. The risk is pages nobody asked for and claims nobody can check, which AI makes cheap to produce at scale. The worth-it verdict prevents the first and the fact gate prevents the second, and a person reads every page for voice before it publishes. An invented statistic is the fastest way to lose an engine's trust in a domain, which is why the content standard behind this library forbids them outright.
The myth piece on AI content being penalised examines the belief in full.
What stays gated, and can it stay gated forever?
Publishing stays at Level 0 by default: every page is approved by the manager before it goes out, and most teams keep it that way permanently, because the FTE's value is in the verdicts, the sourced drafts, and the measurement rather than in unattended posting. The manager can promote a narrow class, factual refreshes of pages from named sources for instance, under the Promotion Standard from a clean record, and step it back at any time. Rewrites of pages that already rank, which the verdict's second question surfaces, are usually the highest-value work in the first quarter and are gated like everything else.
The CMO persona piece describes the publishing decision from marketing leadership's seat.
Can a citation be guaranteed?
No, by anyone. Nobody controls what an AI engine cites, and the engines publish no mechanism to buy or force a citation. What can be done is the documented work, answer-first pages with checkable claims and named entities on questions people actually ask, and the honest measurement of whether it worked, as a dated range per engine. Tactics that claim to force citation, hidden text, mass-generated variants, fabricated statistics, fake citations, are not documented to work and several violate engine guidelines, with domain-level consequences. A vendor offering a guarantee is offering a sales document.
The myth piece on guaranteed AI visibility sets out how to check any claim with a prompt set of your own.
What about traditional SEO, and what about our agency?
The same discipline serves both: a page that answers first with checkable claims ranks in search and is cited in AI answers, and the console movement sits beside the citation range in the Performance tab. Technical SEO, site health, structured data at the site level, and digital PR remain the work of people, at your agency or in-house, and the FTE's pages stand on that foundation. The migration guide from an SEO agency describes what the agency keeps, strategy, technical audits, flagship pieces, and what the FTE takes, the daily measured loop.
The whitepaper on the AI search visibility playbook gives the whole model in one place.