How Bevizio measures AI visibility
Bevizio asks each AI engine the questions your customers might ask, many times a week, and counts how often your product or business is named, who is named instead, and which sources the engine uses. This page says exactly how each engine is asked. The table is generated from the live settings, so it reflects what Bevizio does today.
The engines
| Engine | How the answer is collected | Asks per question per week |
|---|---|---|
| ChatGPT | Captured from the engine's web interface, through a data provider | 3 |
| Claude | The engine's API, through a data provider | 2 |
| Perplexity | The engine's API, through a data provider | 1 |
| Gemini | Captured from the engine's web interface, through a data provider | 1 |
| Google AI Overviews | Read from Google search results (the AI Overview box, when Google shows one) | 1 |
That is 8 asks of every question each week. Engines whose answers vary more from one ask to the next are asked more often. Answers are collected by a third-party data provider, with a location and language set for each brand (the United States and English unless the brand’s country setting says otherwise). For local businesses the town is part of the question. We do not claim this is authorised or endorsed by any of these engines, and Bevizio is not affiliated with any of them.
For local businesses
A local business is asked a different set of engines, still 8 asks of every question each week: ChatGPT 3 times, Gemini twice, Google AI Mode twice and Claude once. Perplexity and Google AI Overviews are not used for local businesses. A question is one service in one town, and the town is written into the question. ChatGPT, Gemini and Claude are collected as the table above says; Google AI Mode, which is used only for local businesses, also comes through the same data provider.
What the numbers are
Every number is a count of real answers, for example “named in 9 of 60”. AI answers differ from one ask to the next, so a count is a sample of a variable system, not a fixed score. A change is only reported when it is bigger than normal variation. When an engine gives no answer, that answer is left out rather than counted as a miss.
An answer counts as “named” in one of two ways: your name (or an alias you have added) appears in the answer text, with accents, spacing and hyphens ignored; or your own site is cited as a source (or its address is written in the answer) and the text does not contain your name. A name that appears only in the title of a page the engine cited is not counted. Next to every named count your dashboard shows how many answers fall into each, so you can see how many of them have your name in the answer itself. Where the full answer text is no longer kept (it is kept for 12 months), an answer can be counted as named while the name is not in the shorter text that remains; the dashboard shows those separately rather than counting them as either kind.
Now and then an engine fails to return an answer. Bevizio tries again, up to four attempts over about eight hours. If it still fails, that answer is left out and marked “couldn’t check” in your dashboard, so a count can rest on a few fewer answers than the usual eight asks a week.
A fair test, by design. Starting questions are written by an AI from your site and category and never contain your product’s name or domain, so an engine can’t name you just because we did. Edit or replace any of them.
The engines collected through their API are asked through these models: Claude (claude-haiku-4-5), Perplexity (sonar). Each of those answers is capped at 1,200 tokens (roughly 900 words), so a very long answer can be cut off before a name that sits near its end. Engines collected from their web interface have no such cap.
The answers can differ from what a person sees in the apps, which vary by account, location, model version and time of day. Where an engine’s collection method changes, your chart shows a marker on that date, and change alerts for that engine pause for four weeks so the switch is not mistaken for a real change in your visibility.
See it for your own product or business
Both free checks need no account. The check for products and websites asks ChatGPT, Perplexity and Google AI Overviews, and the check for local businesses asks ChatGPT, Gemini and Google AI Mode. A paid plan tracks your questions every week: Growth is $29 a month for 20 questions and Pro is $79 a month for 75. Both start with a 14-day trial that takes a card and becomes a paid plan unless you cancel in Settings. The plans are on pricing.
Questions about a number
Every answer behind a count can be opened in your dashboard, with the engine and the date it was asked. If something looks wrong, write to hello@bevizio.com. Who runs Bevizio is on the about page.