AI Search Studies

Almost every large study of AI search visibility is funded by a company that sells AI search visibility software. That is not a disqualification and it is not a secret — the vendors say so — but it does decide what gets measured. A platform that sells prompt tracking measures prompt sets. A platform that sells backlink data measures links. The question each page asks first is what the study could have found had it looked somewhere else.

The recurring design problem is the denominator. None of the AI surfaces publish a query universe, so a study has to choose its own prompt list, and every percentage it reports is a share of that list. Two studies can be internally correct and disagree by a multiple because they bought different lists. Where a figure inherits that limit, the page says so.

The second recurring problem is direction. These are cross-sectional measurements without controls, so a variable that accompanies visibility may produce it, follow from it, or share a cause with it. The studies themselves are usually careful to write associated with. The write-ups that follow them usually are not, and the gap between those two registers is where most bad AI search advice comes from.

Not every study warrants a page of its own. For the wider body of work this industry quotes — who published it, what each headline number is actually a share of, and where two vendors measuring one fact disagree by a factor of two — see the AI search research index.