Measurement

How to Measure AI Visibility: Metrics, Tools and Methods

How to measure whether ChatGPT, Perplexity, Google AI Overviews and other AI engines cite your brand: the metrics that matter, the data sources and a repeatable process.

AI visibility is how often AI-generated answers cite or mention your website when people ask questions in your category. It cannot be read from a ranking report, because AI answers vary by wording, user and day. It is measured by sampling real answers, tracking AI referral traffic and checking that AI crawlers can reach your pages.

The metrics that matter

MetricWhat it shows
Citation rateShare of sampled questions where your site is cited as a source.
Mention rateShare of answers that name your brand, cited or not.
Retrieval rateShare of questions where your pages appear in the results the AI reads.
Share of citationsYour citations as a percentage of all citations across the sample.
Competing domainsWhich sites win the citations you miss.
AI referral trafficVisits from chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com.

Four data sources

1. Live answer sampling

Ask a fixed set of real customer questions with live web search and record which sources each answer cites. Use both unbranded questions (discovery) and branded ones (reputation). The free AI visibility checker does this automatically.

2. Analytics referrals

In your analytics, filter referral traffic by AI platform domains. This shows real visits, though some AI apps do not pass a referrer.

3. Search Console

Google includes clicks and impressions from AI Overviews and AI Mode in the Web search performance report. Watch question-style queries in particular.

4. Server logs

Requests from OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-SearchBot and Claude-User show which AI systems are reading your pages and how often.

A repeatable monthly process

  1. Keep a fixed list of 10 to 20 questions that matter to your business.
  2. Sample answers monthly and record citation rate, mention rate and competing domains.
  3. Track AI referrals and crawler activity alongside the samples.
  4. Improve the pages that should answer each question, then re-test.

Common measurement mistakes

  • Treating one answer as a ranking: answers vary, so use a sample.
  • Tools that report a "ChatGPT rank" without showing the answers they observed.
  • Ignoring branded questions: AI may describe your brand using third-party sources instead of your own site.

Frequently asked questions

Can I see my ranking in ChatGPT?

No tool can see a private ranking. You can measure how often sampled answers cite or mention you, which is the meaningful metric.

How many questions should I test?

Start with 8 to 20 real customer questions, covering both unbranded and branded intents, and keep the list fixed so results are comparable over time.

Why do results change between tests?

AI answers depend on live search results and model sampling, which change daily. Look at trends across several tests.

Check your own site: run a free GEO score check (22 checkpoints with fixes) or test whether AI answers cite you with the AI visibility checker.