How to Measure AI Visibility (and Prove AI Recommends You)
How to track whether ChatGPT, Gemini, Google AI, and Perplexity actually find and recommend your business, using evidence instead of guesses.
The short answer
To measure AI visibility, track three things: whether AI crawlers reach your site, whether the engines name you when you ask the questions your customers ask, and whether that turns into calls. Check your server logs (or a tool) for visits from bots like GPTBot, Google-Extended, PerplexityBot, and ClaudeBot. Then run your real customer prompts through ChatGPT, Gemini, Google AI, and Perplexity on a schedule and record whether you are mentioned, where, and against which competitors. Finally, tie any lift back to calls and booked jobs. Because answers vary between sessions and you cannot control citations, treat each check as a sample and label how confident you are, rather than reading one good answer as proof.
The three layers worth tracking
AI visibility is not a single number. Measure it in layers, from can the engines read you, up to did it produce a customer.
- Access: AI crawlers (GPTBot, Google-Extended, PerplexityBot, ClaudeBot) actually reach your pages. Check this with our free AEO readiness checker.
- Mentions: when you run real customer prompts, the engines name you, and you can see where you stand against competitors.
- Outcomes: calls and booked jobs that follow, the only number that pays the bills.
How to test prompts honestly
Write down the exact questions your customers ask (who fixes furnaces near me tonight, best rated roofer in my city) and run them across the major engines on a regular schedule, not just once. Record the result and a confidence level, because the same prompt can return different businesses in different sessions.
Watch for the difference between a one-off mention and a consistent one. Being named once is encouraging; being named repeatedly, across engines and over weeks, is the signal that matters.
Turn measurement into a baseline
Start with a snapshot so you have something to compare against. Our free Local AI Visibility Check runs your real near-me questions and shows whether AI can find and recommend you today, and where you are missing.
From there, anything you report should resolve to evidence: a logged crawl, a captured answer, a tracked call. That is the honest standard, and it is the opposite of a vanity percentage in a slide.
Key takeaways
- Measure AI visibility in three layers: crawler access, mentions in real prompts, and the calls that follow.
- Test your actual customer questions across engines on a schedule, not once, and record a confidence level.
- A consistent mention across engines and weeks matters far more than a single lucky answer.
- Tie every reported result to evidence: a logged crawl, a captured answer, a tracked call.
Frequently asked
Why do I get different answers each time I ask an AI?
Generative engines are probabilistic and pull from changing sources, so the same prompt can name different businesses between sessions. That is why you sample on a schedule and track consistency rather than trusting one result.
How do I know an AI crawler visited my site?
Look in your server or hosting logs for user agents like GPTBot, Google-Extended, PerplexityBot, and ClaudeBot, or use a tool that surfaces them. If they are not visiting, that is the first gap to fix.
What is a realistic thing to measure first?
Start with a baseline of whether you are named for your top customer questions today. Run a free check, then re-check on a schedule and compare.