One AI reading the answers isn’t enough.
Ask one AI to count who another AI recommended and it will miss names, count “Acme” and “Acme Plumbing” as two businesses, and give you credit for a company that just shares your name. We built checks so those mistakes get caught before they reach your report.
We publish our test results. When we reviewed AI visibility tools in September 2026, we found none that publish how accurate their mention detection is.
Owen designs the questions for each Full Audit and walks you through the results. The counting is done by the checked system on this page, not by hand. Owen spent years in critical care, where you check twice before you act. That habit is built into every report.
Our test results, with the fine print.
We test the system on real AI answers where we already know the right answer, and we keep the tests separate from the answers we tuned it on.
Tested on 480 real AI answers whose labels people verified.
It found 98.3% of the times a business was actually named.
On answers it had never seen, labeled blind by two other AI models, for questions that don’t name the business. This is the Snapshot’s rule.
Same test. Separately, the rival screen kept 14 of 14 known non-businesses, like certifications and directories, off the rival list, and found 89% of the real businesses.
Tested on answers about cannabis shops, wealth managers, a construction company and software firms. Testing on home-services answers is next.
Named. Picked. Pages. Plan.
The $200 Full Audit asks about 100 buyer questions (30 buyer needs, each asked two or three ways, plus about ten that name your business) on five AI tools: ChatGPT, Google Gemini, Perplexity, Google AI Overviews and Google AI Mode. That’s about 500 answers.
You see where you show up, question by question. You see who AI picks instead of you and what sits behind it, what AI says about you, the pages to get listed on, anything blocking AI from your site, and a 90-day plan. Every answer is on an evidence page, so you can check any number yourself.
The $20 Snapshot asks about 25 questions on ChatGPT, Google AI Overviews and Google AI Mode, with the same counting rules, fully automatic.
Get your Snapshot, $20 →Ten ways we check an audit.
Most apply to both the Snapshot and the Full Audit; where one applies only to the Full Audit, the card says so. If you find a gap, tell us.
We only count a business when a visible form of its name is in the answer. No guessing from context.
Is this your business or another one with the same name? Is it offered as an option, picked as the best, or warned against?
When something is unclear, a second AI model reviews it. In the Full Audit, if the two disagree on whether you were recommended, a third model decides. In the Snapshot, a disagreement is shown as “unclear” and left out of the count.
When we can tell a business with your name is a different company, it isn’t counted as you. A dispensary that shares its name with a cash-advance app is not that app.
A business AI names that you didn’t list is screened before it shows up as a rival. Competitors you list are tracked directly.
You and every competitor are counted on the same answers, with the same rules for mentions and recommendations.
Every buyer need is asked in two or three wordings. In the Full Audit, one question in ten is asked again to see if the answer holds. In earlier testing, repeat answers agreed 91% of the time.
If an AI tool fails to answer, that question is left out and the report says so. If Google shows no AI answer at all, we record that too. An AI tool that answered fewer than 80% of the questions isn’t scored.
Before each report, 60 answers we already verified are judged again. If any mention changes, the run stops before a report is made.
Automatic checks hold a report with problems they find, like a quote that isn’t word for word, a number without its “out of”, or a cause-and-effect claim.
See where you stand.
The $20 Snapshot uses the same counting rules and emails you who AI names instead of you and what to do next.