How to measure and report AI visibility: metrics, tools and template

You measure AI visibility by asking a fixed set of customer questions repeatedly in AI assistants such as ChatGPT, Perplexity and Gemini, and tracking in what share of the answers your business is mentioned or linked as a source. That percentage is the core. Share of voice against competitors and indirect signals such as branded searches and AI traffic in GA4 complete it. One thing to know up front: AI answers differ a lot from one time to the next, so a single measurement says little. This article shows which metrics count, how to measure reliably, which tools you have and how to turn it into a report for your client or manager.

Why can’t you see AI visibility in Google Analytics?

Because AI assistants do not pass on impressions. Google Search Console shows how often you appear in Google, but ChatGPT, Perplexity and Gemini have nothing comparable. For Google’s own AI answers there has been a report in Search Console since August 31, 2026 with impressions from AI Overviews and AI Mode (Search Engine Journal), but without clicks and without showing whether your brand was in the text of the answer.

So whether you are mentioned has to be measured in the answers themselves. Your analytics only tells you what happens after someone clicks through; how to separate that AI traffic in your stats is on GEO Analytics and in tracking ChatGPT traffic in Google Analytics.

Why one measurement says nothing: noise in AI answers

AI assistants write every answer anew, and that creates a lot of noise. SparkToro and Gumshoe had about 600 volunteers ask 12 identical questions to ChatGPT, Claude and Google AI in January 2026, 2,961 times in total. The result:

• The chance that ChatGPT or Google AI gave the same list of brands twice for the same question was under 1 percent.
• The same list in the same order appeared about once per 1,000 answers.
• Still, some brands appeared in 60 to 90 percent of answers for the same intent. So how often you are mentioned is a usable measure.

What does that mean for your measurement?

1. Measure every question several times. Rand Fishkin of SparkToro advises asking every question 60 to 100 times and taking the average. How many measurements are needed exactly, he calls an open question.
2. Report percentages over a period, not one day’s answer.
3. Do not rely on positions. Whether you are named first or third differs almost every time.
4. Do not call a small change a trend. A rise of a few percentage points can simply be noise. See whether a movement holds for several weeks.

That is why GreatCompany measures daily. A monthly number then rests on a series of measurements instead of one snapshot.

Which metrics matter?

The most important metric is your mention rate: in what share of the answers to your questions are you mentioned. The rest helps explain that number.

Metric	What it measures	Calculation
Mention rate	How often your brand appears in the answers	Answers with your name / all measured answers × 100
Citation rate	How often your site is linked as a source	Answers with a link to your site / all measured answers × 100
Share of voice	Your share against competitors	Your mentions / mentions of you plus your competitors × 100
Question coverage	For which questions you appear and for which not	Your mention rate per question; questions without a mention are your gaps
Accuracy	Whether the AI talks about you correctly	Manual check of prices, services, location and tone

Two notes. First, a mention is different from a citation: ChatGPT can recommend your business without linking to your site, and Perplexity can use your page as a source without naming your brand in the text. Keep them separate. Second, there is no general benchmark for a good rate. It depends on your market and on how broad your questions are. Compare your number with your own previous periods and with your competitors on the same questions.

Indirect signals: what AI visibility brings in

Besides the answers themselves, you measure the consequences in tools you already have:

• Branded searches (Search Console). People who see your name in an AI answer often look you up in Google afterwards. A rise that does not come from a campaign can point to AI mentions. See connecting Search Console to AI visibility.
• AI traffic (GA4). Visitors from chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com come in as referral traffic. Some arrive without a referrer, for example from apps, and fall under direct traffic. How to set that up is in tracking ChatGPT traffic in GA4.
• Crawler visits (server logs). Visits from GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot show that AI platforms fetch your pages. ChatGPT-User and Perplexity-User fetch a page because a user is asking a question right then; that is the strongest signal you are used in answers. More in how AI crawlers read your site.
• Inquiries. Add “How did you find us?” to your contact form, with ChatGPT or another AI assistant as an option. That catches what analytics misses.

Which questions do you measure?

Measure the questions your customers ask before they buy, in their own words. A good set has 20 to 50 questions and mixes four kinds:

• Category: “Which accounting software is good for a freelancer?”
• Comparison: “What is the difference between X and Y?”
• Problem: “How do I file quarterly estimated taxes without an accountant?”
• Local: “Good heat pump installer in Columbus”

Keep the set fixed, otherwise you compare apples and oranges. Adjust it per quarter, and note in your report when you did. More in which prompts to track.

How to measure it: from free to automatic

You can start free with manual testing and proxy metrics. A tool only becomes necessary when you want to measure more often and across more questions than you can by hand.

Manual testing (free)

1. Take 10 to 20 questions from your set.
2. Ask each question a few times in ChatGPT, Perplexity and Gemini, each time in a new conversation. Use a temporary chat or an account without history, so earlier conversations do not color the answer.
3. Note per answer: are you mentioned, are you linked, which competitors are in it, and is what it says about you correct.
4. Calculate your mention rate per month.

Pro: it costs nothing and you see the answers with your own eyes. Con: it quickly takes a few hours a week, and with a handful of measurements per question the noise stays large.

Proxy metrics in tools you already have

Search Console, GA4 and your server logs give the indirect signals above. A Looker Studio dashboard that puts branded searches and AI traffic side by side takes an afternoon to build.

A prompt tracker

A prompt tracker asks your questions automatically and keeps track of the results. When choosing, look at:

• Measurement frequency: how often is each question measured? More measurements mean less noise.
• Platforms: which AI assistants are measured?
• Competitors: do you see share of voice per competitor?
• History: can you see trends over months?
• Sources: do you see which pages the AI uses as a source?

GreatCompany measures every day whether ChatGPT, Gemini and Perplexity mention your business, next to your Google data. A comparison with other tools is in the overview of GEO tools, and the difference with classic rank tracking in prompt tracker vs rank tracker.

How often to measure and review what

What	How often to measure	How often to review	Source
Mentions in AI answers	Daily or as often as possible	Monthly, as an average	Prompt tracker or manual
Share of voice	Along with the mentions	Monthly	Prompt tracker or manual
Branded searches	Continuously	Monthly	Google Search Console
AI traffic and inquiries	Continuously	Monthly	GA4 and your CRM
Crawler visits	Continuously	Monthly	Server logs
Question set and content	Per quarter	Per quarter	Manual

Reporting AI visibility to your client or manager

A good report starts with the result, compares with the previous period and translates the numbers into what a decision maker already knows: brand awareness, traffic and inquiries. That is harder than with SEO, for three reasons.

1. The metrics are new. Your reader knows rankings and organic traffic, but not a mention rate or share of voice in AI answers. You have to explain them the first time.
2. The impact is indirect. With SEO the chain runs from ranking to click to inquiry. With AI visibility it often runs from mention to branded search to inquiry. Make that chain explicit.
3. There is little history. With SEO you compare with last year; with AI visibility you often have only a few months of data.

The template: six parts

1. Summary (one page, 30 seconds)

• Key number: mention rate this month, with last month next to it. For example: “Mentioned in [x]% of answers, was [y]%.”
• Branded searches: change against the previous period.
• Main result: a concrete question with buying intent where you are now mentioned.
• Priority for next month.

A traffic light (green, amber, red) per key number works well for decision makers.

2. AI visibility in detail

• Mention rate per month as a line chart.
• Share of voice against your top three to five competitors as a bar chart.
• A table per platform:

Platform	Mention rate	Previous period
ChatGPT	[x]%	[y]%
Perplexity	[x]%	[y]%
Gemini	[x]%	[y]%
Google AI Overviews	[x]%	[y]%

3. What we did

Per action (page rewritten, FAQ added, mention on an industry site) the effect on the questions that action was about. Be honest: only link an effect to an action if the change is larger than normal fluctuation and holds for a few weeks.

4. Competitors

• The five questions where competitors are mentioned and you are not.
• The five questions where you are mentioned and they are not.
• The development of share of voice over the last three months.

5. Business impact

This is the most important page for leadership. Put branded searches, AI traffic from GA4, inquiries from that traffic and the answers to “How did you find us?” together. Avoid calculating an estimated “media value” of AI mentions: there is no reliable method for it, and a decision maker sees through it quickly. Show for which questions with buying intent you are now mentioned; that is convincing enough. More on calculating return in the ROI of GEO.

6. Plan for the next period

The three most important actions, what you expect from them and what they need.

Tips for a report that gets read

• Start with the result, not the method. A manager wants to know whether things are going well. The method belongs in an appendix.
• Always compare with the previous period. “From 23 to 28 percent” says more than “28 percent”.
• Be open about noise. Explain that you report averages over many measurements because AI answers differ every time. That keeps a bad day from causing panic, and makes your numbers more credible.
• Use screenshots as illustration. A screenshot of ChatGPT recommending your business makes it tangible. Do not present it as proof: it is one answer out of many.
• Explain the metrics the first time. For example: “The mention rate is our market share in AI answers.”
• Six pages at most. More does not get read. Extra data goes into an appendix.

How often to report

Audience	Frequency	Focus
Leadership	Monthly	Summary and business impact
Marketing team	Every two weeks	Metrics and actions
Client (agency)	Monthly	Full template
Wider stakeholders	Per quarter	Strategic review and plan

Start today

You do not have to wait for a tool. Pick ten questions, measure them by hand a few times a week and put your branded searches and AI traffic next to them. After a month you have a first baseline.

Read also

• Which prompts to track for AI visibility
• Tracking ChatGPT traffic in GA4
• Prompt tracker vs rank tracker
• GEO glossary