What does the myth say?
That there is a technique, a tool, or a vendor that can ensure your brand appears when an AI engine answers a question in your category, and that visibility in AI search is therefore something to buy with a guarantee attached. Marketing leaders under pressure to be in the AI answers hear this claim often, and the ones who accept it spend a quarter on a promise nobody could keep.
The examination is about what is controllable, what is measurable, and what should be refused.
Where does the belief come from?
From the history of search, where paid placement and, for a time, manipulable ranking factors made guarantees seem plausible. From vendors who have relabelled old tactics as AI optimisation and attached a promise. From the genuine anxiety of watching AI answers replace clicks. And from the fact that some practices do measurably help, which makes it easy to slide from helps to guarantees.
The belief is an old habit applied to a new surface that does not work the same way.
What is true in it?
That AI engines cite sources, that some pages are cited far more than others, and that the difference is not random. That pages which answer the question first, in a self-contained chunk, with claims that can be corroborated elsewhere and entities named explicitly, are cited more often; the engines' own guidance and observed behaviour agree. That structured data, freshness, and being the source other sources cite all help. That a company which does this consistently for a quarter usually sees its citation presence move. Visibility is influenceable, and the practices are known.
What is not true is that anyone can promise the outcome for a given query on a given day.
What does the mechanism actually do?
A Search Visibility Digital FTE on ftes.ai samples a fixed prompt set, the questions buyers actually ask, across search and AI engines every week, with dates, and records whether the site is cited. It proposes topics with a worth-it verdict from that data, drafts through the fact gate so every claim is checkable, structures each page answer-first with the entities named, and publishes on approval through the site's own route. It then measures again and reports citation presence per engine as a range over the month, with the sample dates, and marks as unmeasured anything it cannot see. It never reports a guarantee, because it has nothing to report one from.
The whitepaper on AI search visibility in the library sets out the practices and the measurement in full.
- Sample: fixed prompt set, per engine, weekly, dated
- Do: verdicts, fact-gated drafts, answer-first structure, entities named
- Report: ranges with dates; unmeasured labelled
Which tactics should be refused outright?
Anything that claims to force a citation: hidden text aimed at engines, mass-generated pages targeting every phrasing, fabricated statistics to seem authoritative, fake reviews or citations, and content injected to manipulate an engine's retrieval. None is documented to work reliably, several violate published engine guidelines, and an uncorroborated figure or a manipulation pattern can lower an engine's weighting of the whole domain. A Search Visibility FTE will not propose them, and the fact gate blocks the fabricated-statistic version mechanically.
The honest rule is that a tactic which would embarrass you if the engine described it publicly is one to refuse.
How can you check a claim for yourself?
Ask any vendor, including this one, for the prompt set, the engines sampled, the sample dates, and the result as a range. Ask what is unmeasured and see whether the answer is nothing. Ask what happens if a target query is not cited after a quarter; the honest answer is that the record shows what was tried and the next quarter's plan is revised, not that a guarantee is invoked. Then sample ten of your own buyer questions across three engines yourself, this week and in four weeks, and compare. You will have a baseline no vendor gave you and a sense of how much movement is realistic.
A range with dates is a measurement. A promise is a sales document.