The figures that were on this page have been withdrawn. They were produced by a method we have since found to be wrong, so they are not being shown while the dataset is re-measured.
What was wrong. To score a site we ask an answer engine the questions its buyers would type, then check whether that site gets named. Until 26 August 2026 those questions were written by a model that had just read the site's own homepage, so it phrased them in that site's own vocabulary. Each question then goes to a live web search, and a page is the best match on the web for its own wording, so the site came back top and was scored as visible.
The effect is systematic and it always flatters: it measured how distinctive a company's marketing language is, not how visible it is. On one site the same audit read 4 of 6 under the old method and 1 of 6 under the corrected one.
What we changed. Buyer questions are now generated from the category alone, by a step that never sees the page, in the plain words a stranger shopping in that market would use. Any question that echoes a distinctive phrase from the site's own headline is discarded.
Where it stands. 0 sites re-measured under the new method, 44 still to go. Figures return here once at least 12 are done, and the numbers will be lower than the ones this page used to show.
It was too high, and by an amount we cannot state as a single number, because the size of the error depends on how distinctive your own website's language is. A company whose homepage uses the plain words of its market was barely affected. A company with a strong, unusual positioning line was flattered most, because the questions we generated echoed that line back and almost nobody else could have been the answer to them.
Re-run the audit and you will get the corrected number. If you paid for a report built on the old method, tell us and we will re-run it and re-issue it at no cost.
The test only means something if the question is one a stranger would actually type. For an affiliate marketing agency the fair questions are "how do I find a good affiliate marketing agency" and "what is the best affiliate marketing company" — plain, generic, and winnable by any firm in the market. The question we generated instead was "what are the best affiliate marketing services for challenger brands", which is that company's own homepage phrasing. Being the answer to your own tagline is not visibility, it is an echo.
The same site scored 4 of 6 on the old questions and 1 of 6 on the new ones, and on the corrected run six genuine competitors appeared in answers where previously none had.
Scores from the old method and the new one are not the same measurement, so averaging them produces a number that describes neither. Every stored result is now stamped with the method that produced it, and this page computes only from rows measured the current way. That is why it shows a count instead of an average today, and why it will start showing figures again on its own once enough of the dataset has been re-measured. Nothing here is being hand-edited back into agreement.
Published 2026-08-26. We would rather show you this than a number we no longer believe. Run the corrected audit on your own site.