How we measure AI search visibility

1. We start with your buyers' questions

Not keywords. Real questions, with the details people actually type: the place, the problem, the deal-breaker. “Best implant dentist in Houston for nervous patients”, not “dentist Houston”. A full audit maps 100+ questions across every service, location and way of asking, and you approve the list before we run it.

2. We ask every engine, more than once

Each question goes to ChatGPT, Perplexity, Gemini, Claude and Google's AI Overviews, five times per engine, in fresh sessions with no chat history. 100 questions across five engines, five runs each, is 2,500 answers. We keep every one of them, word for word, with the sources the engine cited.

3. What counts as being named

• Named: your business appears in the written answer.
• Named first: you're the first business the answer recommends.
• Cited: the answer links to your website as a source.
• On the map only: you show up in a map card but the written answer skips you. We track this separately, because it usually points to a fixable listing problem.

4. The numbers in your report

5. Why we show a range, not one number

If you're named in 9 of 48 answers, the honest result isn't “19%”. It's “about 19%, likely somewhere between 11% and 30%”, because a different set of runs would land a little differently. We calculate that range from how many answers we ran and show it next to every headline number. The more runs, the tighter it gets. Most tools in this space show a single number with no range at all.

6. Where AI gets its information

Every source an engine cites gets sorted, so you can see which ones you control and which ones you need to get onto:

• Your own website
• Competitor websites
• Directories and listings
• Review sites
• Forums and communities
• News, guides and best-of lists
• Map listings

Fixes target the sources the engines actually used for your questions, not a generic checklist.

7. How we put a dollar figure on it

Monthly searches for your services × the share of searchers who ask AI × the share who book when AI names a business × your average customer value × the share of answers that leave you out.

Every input is labelled as sourced or an assumption, and you can change any of them inside your report. It's an estimate, and we say so on the page. Its job is to show which fixes are worth doing first.

8. Limits we're upfront about

• Answers can differ for someone logged in with a long chat history. We measure clean sessions, the closest thing to a new buyer.
• Engines update their models. That's why we re-run every question every 30 days and report the change, not just the snapshot.
• Every question names the market the way your buyers ask it, so local results are about your area, not ours.
• We never promise a ranking or a first position. Nobody controls that. We promise to measure honestly and to move the numbers we target.