How we measure AI visibility
A lot of what is sold as AI visibility measurement is sampling presented as a percentage. Here is what we measure, how, and where the limits are.
What can be measured directly
Some AI visibility shows up in data you already own. We start there, because nobody has to take our word for it.
| Source | What it shows | Its limit |
|---|---|---|
| GA4 referrals | Visits from AI answers that linked to you: chatgpt.com, perplexity.ai, copilot.microsoft.com, gemini.google.com | Misses every answer that named you without a link |
| Bing AI Performance | How often Copilot cites your content | Microsoft only, and in beta |
| Search Console | Google traffic, with its AI features counted inside the web totals | Not broken out separately |
| Branded searches | People searching your name, often after meeting it somewhere | Indirect. Many things cause it |
What has to be sampled
Whether ChatGPT, Claude or Gemini names you in an answer without a link cannot be read from any report. The only honest method is to ask, the same way, on a schedule.
- A fixed list of questions. Twenty to thirty that your buyers really ask, in their words. The list does not change between rounds.
- Fresh sessions. No conversation history, so nothing personal colors the answer.
- The same tools, the same day each month.
- Four things recorded: whether you were named, who else was, which sources the answer leaned on, and whether anything said about you was wrong.
The sources column is usually the most useful one. It shows which directories and publications shape how your category gets described, and those are places you can act on. Our free AI visibility tracker is a basic version of this step.
What cannot be measured
- Your share of all AI answers in your category
- What a specific person saw when they asked
- Why a model chose one source over another
Tools that report an "AI visibility score" are running samples of their own. Some are useful. Ask how many questions, how often, from where and across which tools before trusting the number.
How it appears in a report
Every figure we report carries its source, its date range and any caveat that applies. Samples are labeled as samples. The standard we hold reports to is written down, so you can hold us to it.
Want a baseline you can trust?
We will build your question list, run the first round, and show you exactly how every number was made.