Which AI search numbers survive checking
Four statistics about AI search circulate constantly in marketing posts. We read the primary source for each one. Two of the four do not say what they are quoted as saying. A third is a survey of opinion being cited as measurement, run by an agency selling the service it asked about. One survives intact.
Why check a number at all?
A statistic with a company name attached reads as evidence. Most of the time nobody opens the source, and the number travels anyway. We check them because a page about being accurately described by machines is the worst possible place to publish something that is not true.
Which numbers did not survive?
A pair of figures about impressions, quoted as a matched set, appear in no source in the form claimed. A keyword search returns a real study whose number looks like a match and is not: that study reports 71 million clicks, a sample size, not a percentage. The digits matched; the claim did not.
That is the trap worth naming. A citation is verified by reading the page, never by finding the number.
A fifth number, checked in October 2026
A 28-page report circulated as Gartner's AI Hype Cycle 2025 states that 57% of organizations admit their data is not AI-ready. The report is not published by Gartner. It is published by another company, whose name appears in the footer of every page.
Gartner's own press release, Lack of AI-Ready Data Puts AI Projects at Risk, dated 26 February 2025, says this instead: 63% of organizations either do not have or are unsure if they have the right data management practices for AI, from a survey of 1,203 data management leaders taken in July 2024.
Three things moved. The number went from 63 to 57. The population went from data management leaders to organizations in general. And “either do not have or are unsure” became “admit”, which turns uncertainty into a confession.
The same report carries a second figure, about average spend and executive satisfaction, that we could not find on Gartner's site in any form. We are not calling it false. We are saying we could not source it, which is a different thing and the reason it is not quoted here.
What about the ranking-factor percentages?
A widely-shared table rates thirteen factors for AI visibility, with brand mentions at 94% and backlinks last at 3%. The methodology block on the source page describes it plainly: an online survey of 500 marketers and business owners, identifying factors by perceived importance.
So it is a record of what marketers believe drives AI visibility, not a measurement of what does. Published as "what drives AI visibility" it misstates its own source. It was also run by an agency that sells the service it surveyed, which is not disqualifying but is worth saying out loud.
Which one holds up?
Ahrefs analysed 300,000 keywords and found AI Overviews correlated with a 34.5% lower average click-through rate for the top-ranking page. That is a measurement, on a stated sample, published with its method. Pew Research separately found people clicked a traditional search result in only 8% of visits when a Google AI summary appeared, against nearly 15% when none did. BrightLocal found 40% of consumers actively use generative AI when searching online.
What should you do with a statistic someone quotes at you?
Ask what the sample was and who paid for it, then open the source. If the number turns out to be a sample size, a survey of opinion, or absent entirely, you have learned something more useful than the number.
Questions
Which AI search statistics are reliable?
The Ahrefs analysis of 300,000 keywords, the Pew Research click-through finding and the BrightLocal adoption figure each state their method and sample. A widely-shared table of thirteen ranking factors is an opinion survey of 500 marketers, and a commonly quoted pair of impression figures appears in no primary source in the form claimed.
Why does a survey of marketers not count as evidence?
It records what people believe influences AI visibility, not what does. Published as a list of ranking factors it misstates its own methodology, which describes perceived importance.
How do I check a statistic myself?
Open the primary source rather than a page quoting it, and read the methodology block. Check whether the number is a percentage or a count, and who funded the work.