How to measure AI visibility (the honest, free way)

Most people try to measure AI visibility with the old SEO ruler, and it does not fit. There is no rank to check. An AI answer either names your business or it does not, and that answer can change from one month to the next. So the honest way to measure AI visibility is simple. Ask the questions your buyers ask, in the real AI engines, on a schedule, and keep score of who gets named.

The good news is you do not need a paid tool to start. A free monthly prompt test tells you most of what you need to know. This guide walks through that test, the three metrics that actually matter, how to see AI referral traffic in GA4, and why keyword rankings quietly stopped being the answer.

How do you measure AI visibility?

You measure AI visibility by running a fixed set of buyer questions through the main AI engines every month and recording who gets named. This is called a prompt panel, and it is the single most useful thing you can do. It turns a vague feeling ("are we showing up in ChatGPT?") into a number you can watch go up or down.

Why this works. AI engines answer questions, they do not list ranked results. So the only real measure of your visibility is how often an engine picks you when a buyer asks. If you name your 15 questions once and reuse them every month, you get a fair, repeatable read. Change the questions each time and you are measuring noise.

Here is the naming test, step by step. It matches the numbered how-to further down, so you can follow either one.

1. Write about 15 buyer questions. Use the words a real customer would type, like "best HVAC company near me" or "who does emergency water damage repair in Austin". Not your keywords. Their questions.
2. Ask each one in the main engines. ChatGPT, Perplexity, Google AI Overviews, and Gemini. Use a clean or temporary chat so your own history does not tilt the answer toward you.
3. Log who gets named. For each question, note if the engine named you, and which competitors it named instead.
4. Repeat on the same day next month. Same questions, same engines. Now you have a trend.

Two prompts to add to the panel. First, a brand-accuracy check. Ask each engine "what is [your business] known for?" and read whether the answer is correct and positive. If a bakery comes back described as a plumber, the engines lack clear mentions of what you actually do, and that is the work. Second, an attribution prompt. After an answer, ask "why did you give this answer, and list ten things that influenced it." The reply shows you which sources the engine leaned on, so you know where to earn more mentions. A wider read on brand mentions and share of voice sits in its own guide. See how to track brand mentions in AI for the deeper version of the naming test.

What is the full AI visibility measurement stack?

The full AI visibility measurement stack is the monthly prompt panel plus four free data layers that sit on top of it. The prompt panel is the core, because it is the only thing that directly shows who AI names. The four layers add proof from tools you may already have. Here is the whole stack in one view.

Notice how much of this is free. The core of AI visibility measurement is a spreadsheet and a monthly hour, not a subscription. The next four sections walk through each free layer, and honest caveats for each. One caveat holds across all of them. There is no AI search console. No engine gives you a clean dashboard of every AI answer you appear in, so every layer here is a partial read that you stitch together.

How do I track AI referral traffic in GA4?

You track AI referral traffic in Google Analytics 4 by filtering for the AI engines as referral sources. This shows you when an AI answer actually sends a person to your site. Build an exploration or a channel filter that looks for referrals from these hosts.

Two honest caveats about this list. First, there is no official AI channel in GA4, so this filter is our own labeled reconstruction of which hosts count as AI. New engines appear, so keep the list current. Second, consent mode undercounts. Visitors who decline analytics cookies do not get tracked, so your real AI referral number is higher than what GA4 shows.

And a caveat about size. AI referral volumes are small today, and independent data says they will stay small for a while. SparkToro and Datos (Rand Fishkin) reported AI tools send under 1 percent of outbound web traffic. So do not judge AI visibility by referral clicks alone. Treat the trend as the signal, and lean on the naming test as your main measure. For the traffic and evidence side of this in more depth, see how to get cited by LLMs, with evidence.

How do I read Search Console as an AI fan-out detector?

You read Google Search Console as an AI fan-out detector by filtering the Performance query report for long, natural-language questions, roughly 7 words and up. AI engines break one buyer question into many hidden sub-questions, a pattern called query fan-out, and those sub-questions look like full sentences, not short keywords. So the long queries you already surface for are a map of where AI can pull you in.

Always drill down to one query and one page. The big totals at the top of Search Console are close to meaningless on their own. Average position and click-through rate only make sense once you pick a single query paired with a single page. Then the numbers tell a real story.

The impressions clue. Some SEO practitioners read one pattern as an AI signal. A page sitting around position two, with lots of impressions but almost no clicks, can mean an AI answer is taking the click before the visitor reaches you. Treat this as an interpretation, not a hard number. It is a clue worth checking, not proof, and other things can cause the same shape, like a featured snippet or a weak page title. Search Console data is also about a day stale, and staler during Google updates, so give it time before you react.

When tools disagree, trust Search Console. If a paid GEO dashboard and your Search Console impressions tell different stories, weight the impressions. They are real Google data. Many GEO dashboards are modeled estimates, so give the measured number more weight than the estimated one.

What does Bing Webmaster Tools show about AI?

Bing Webmaster Tools now carries an AI Performance report that shows grounding queries and citations. Grounding queries are what people actually searched, and citations are how many times your pages were used in the answer. It is the closest thing to first-party AI citation data that any free tool gives you today.

The scope matters, and it is easy to over-read. This report covers Microsoft Copilot, the AI built into Bing, not ChatGPT and not Gemini. So it is a window into one engine, not all of them. Even so, it is useful, because the questions people ask Copilot are often the same ones they ask elsewhere. To set it up, sign in to Bing Webmaster Tools, import your site from Google Search Console, and open AI Performance in the left menu.

Turn grounding queries into content. The queries you get cited for are a ready-made list of article ideas. If Copilot cites you for a question you never wrote a page about, that is a page to write. This closes the loop between measuring and improving.

Which metrics actually matter?

Four metrics matter for AI visibility, and keyword rankings are not one of them. Here is what to watch, in order.

1. Citation frequency. How many of your buyer questions name you, across all engines. This is the headline number.
2. Share of voice. When you are named, how often is it you versus your top two or three competitors. Being named 1 time out of 15 next to a rival named 12 times is a very different picture than an even split. Pick three named competitors and log them in the same panel, engine by engine.
3. AI referral sessions. The GA4 number above. Small today, but a real trend line.
4. Entity mention velocity. Whether the count of other sites mentioning your business by name is rising. This is the leading indicator, because mentions feed citations.

Why rankings are the wrong metric. People assume that if you rank in Google, AI names you. The data says no. Ahrefs studied 15,000 queries and found only about 12 percent of AI-cited URLs also rank in Google's top 10. That means roughly 4 out of 5 AI citations go to pages that are not ranking at all. Ahrefs sells SEO tools, so read that with care, but the finding cuts against their own product, which makes it more believable, not less. Google rank and AI visibility are two separate scores. Never blend them into one number.

Why mention velocity matters. The same source studied 75,000 brands and found branded web mentions were the strongest correlate with AI-Overview visibility, at a Spearman r of 0.664, about 3x stronger than backlinks. It is a vendor study and correlation is not causation, the authors say so themselves. But it points the same way our own work does. Getting named by AI is won mostly off your own site, through other trusted places mentioning your business. The share-of-voice side of this lives in tracking brand mentions in AI.

What is an AI visibility score?

An AI visibility score is a single rolled-up number for how often and how prominently AI engines name your business across a set of questions. Think of it as one figure that captures your whole prompt panel, so you can say "we went from 3 out of 15 to 8 out of 15 this quarter" instead of reading a spreadsheet.

Why it helps. A score gives you a baseline. Without one, you cannot tell if a change you made moved anything. There is no official, industry-wide AI visibility score yet, so any score is only as honest as the questions and engines behind it. Ours is built from a real prompt panel, not a black box. The free AI-Visibility Audit produces a score for your site so you have a baseline number to beat.

Can I measure AI visibility for free?

Yes. The monthly prompt test costs nothing but your time, and it is the same test we run in a paid audit. You ask your buyer questions in each engine, log who gets named, and compare month over month. GA4, Search Console, and Bing Webmaster Tools are free too. That is the whole stack.

What about paid tools? A category of AI-visibility trackers now exists that automates this at scale, checking hundreds of prompts across engines and charting share of voice for you. They save time once you are running dozens of questions. But read every price and claim with care, because most carry a conflict of interest. Traffic-estimate tools like SimilarWeb report vendor-estimated numbers, not measured ones. Some trackers are run by the same people who sell the wider SEO suite, like Ubersuggest. Newer AI-only trackers such as Profound and Peec are vendor tools with their own pitch, and one tool widely cited for AI citation tracking has an uncertain name in the source we found it in, so we will not print it here until we confirm it. We are working on an honest tools roundup so you can compare them fairly. Until then, the free manual test is more than enough to get a real baseline.

How often should I measure AI visibility?

Measure monthly. AI answers shift as models get updated and as new sources get cited, so the picture from three months ago may already be wrong. A single check is a snapshot. Monthly checks are a trend, and a trend is what tells you whether your work is paying off.

Keep one honest idea front of mind. AI visibility is probabilistic. You are not holding a position the way a number-one ranking holds a spot. You are raising the odds that an engine names you, and those odds move from day to day. An answer might name you 15 hours out of 24 and skip you the rest, and that is still a good result. So measure the trend, not any single answer, and do not panic over one bad check.

Pick a day, like the first Monday of the month, and run the same panel every time. Freshness matters on the other side too. Engines lean toward recently updated, well-sourced pages, so a business that keeps measuring and improving tends to pull ahead of one that set things up once and walked away.

How to measure AI visibility, step by step

This is the full routine in five steps. Run it once to get your baseline, then repeat it monthly.

1. Pick about 15 buyer questions. Write down the questions a real buyer would type when looking for a business like yours, in their words, not your keywords.
2. Run them monthly across the main engines. Ask each question in ChatGPT, Perplexity, Google AI Overviews, and Gemini, using a clean or temporary chat so past history does not skew the answer.
3. Log who gets named. Record whether each engine names you, how many questions name you, and which competitors get named instead.
4. Add the free analytics layer. Filter GA4 for referrals from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, and claude.ai, check Google Search Console for long natural-language queries, and open Bing Webmaster Tools AI Performance for Microsoft Copilot citations.
5. Review month over month. Compare this month to last. More questions naming you, plus rising AI referral sessions, means your AI visibility is climbing.

Once you know your number, the next question is how to raise it. Two pages to read next: how to get cited by AI for the on-page and off-site work, and AI crawler access to confirm the bots can even read your site, because if they are blocked, nothing else you measure will move.

Common questions about measuring AI visibility