The developer API and the consumer app disagree on 76% of brand mentions and 96% of cited sources. We measure all three: the AI models asked directly, Google AI Overviews, and ChatGPT as customers use it, in a logged-in browser. Every number in the report says which test produced it.
"Is Strozzi reliable?" mentions Strozzi 100% of the time by tautology. We report how often AI mentions you when the question doesn't include your name. When it does include your name, that's a sanity check, never the headline.
Ask the same question twice and 40–60% of the answers change. We report intervals, not points. Trends beat snapshots. Anyone promising a specific AI ranking is selling something this methodology won't.
Every fix is a staged change with an approval record and a rollback handle. Nothing reaches your site without your recorded consent.
We ask five AI engines 100 real customer questions in your market and language.
For every weak answer we name the cause and tie it to a specific page or claim.
The platform drafts each fix, you approve it, and every change can be undone.
We ask the same questions again and show which fixes moved the numbers.
SEO is about ranking in the list of links. AI visibility is about what the AI answer says before anyone clicks a link. The measurement and the fixes are different.
No. Ask the same question twice and 40–60% of the answers change. Anyone promising a specific ranking is selling something this methodology won't. We report intervals and trends, not points.
A 12-part report: how often AI finds you, whether it recommends you, what it gets wrong, whether it can read your site, where its sources come from, and how you compare to competitors. Plus a fix plan ordered by impact and effort.
90 minutes end-to-end for the full audit. The free snapshot lands in 30 seconds, enough to tell you whether AI engines can read your site at all.
