How to measure AI search visibility without guessing

One ChatGPT answer tells you almost nothing about whether AI recommends you. Here is how many answers you need, which questions to ask, and what to count.

Mike Wilson · November 1, 2026 · 3 min read

Free, no card, first answers in under a minute.

The usual way to check whether AI recommends your business is to open ChatGPT, type a question a customer might ask, and see if your name comes up.

It feels like a measurement. It mostly isn’t.

One answer is a coin flip

Ask the same question twice and you can get a different list. The models are built to vary, and when they search the web first, what they find varies too. So a single answer tells you what happened once. It does not tell you what usually happens.

Here is the arithmetic. Say AI names you in 4 of 10 answers. The honest reading is not “40%”. It is “somewhere between about 17% and 69%”. That range is just what ten coin flips can do. At 40 of 100 answers the range narrows to roughly 31% to 50%. You need dozens of answers before the number means much, and hundreds before small changes mean anything.

So here is the first rule. Never report a share without the number of answers behind it. If a tool shows you a percentage and not how many answers it came from, assume the range is wide.

Ask the questions buyers ask, not your name

The second mistake is asking about yourself. “What do you know about Acme?” tests whether the model has heard of you. It does not test whether it recommends you. Customers don’t ask about you. They ask about their problem, such as which rain jacket lasts, which accountant does small-business taxes, or which tool tracks brand mentions in ChatGPT.

Write ten to fifty questions the way a buyer would phrase them, before they know your name. Leave your brand out of every one. Then the only way you appear is if the AI chose you.

Count three things, not one

Whether you are named is the obvious number. Two others matter as much.

Position. Being named fifth in a list of six is not the same as being named first. Track the average place you appear when you do.

Sources. When AI recommends someone, it usually leans on pages it found, such as a comparison, a review, a buying guide, or the brand’s own site. Those cited pages are the part you can act on. You cannot argue with a model, but you can make sure the pages it reads describe you accurately, and you can get onto the round-ups it trusts.

When we ran this on our own company, a new one, the answer was humbling and useful. We were named in none of the answers to the ten questions our buyers ask. Four competitors were each named in roughly 35 to 45% of them. And most of what the AI read was the competitors’ own sites and a handful of “best tools” round-ups. That is a to-do list, not a verdict.

Measure the AI your customers use

There is more than one ChatGPT. The app people use on their phones adds its own search and its own instructions. The API that most tools call does not. In one of our tests the app named a brand in 4 of 7 answers and the API in 2, and the app cited the brand’s own site in 5 of 7 answers while the API cited it in none. If you only measure the API, you are measuring something your customers never see.

Then measure again

Answers drift as models update and as the web changes. A number from last month is a number about last month. Check on a schedule, the same questions each time, and look at the trend, not the latest reading.

Our public AI rankings work this way. The same buyer questions go to every engine each week, and every share comes with its range and the number of answers behind it.

The short version

  • Ask many questions, many times. Report how many answers sit behind every number.
  • Ask what buyers ask, without your name in it.
  • Count position and sources, not just mentions.
  • Check the apps people use, not only the APIs.
  • Do it on a schedule and watch the trend.

You can do all of this by hand with a spreadsheet and patience. How to track ChatGPT mentions over time lays out the sheet and the weekly routine. We built Behavior because we wanted to do it every day without the spreadsheet. It asks your buyers’ questions in the ChatGPT and Gemini apps and in Google’s AI answers, and through the APIs of models such as GPT, Claude, and Gemini. Then it shows the share, the range, the position, and the sources behind it.

Run the free check to see what Google AI Overviews and AI Mode say about your brand. A free account adds ChatGPT.

Check your AI visibility free.

See what Google AI Overviews, AI Mode, and one AI model tell your customers, and what to change. Free, no card, first answers in under a minute.