How to Measure AI Visibility in Google Search, ChatGPT and Perplexity

How to measure AI visibility is to count three things together, not one number alone: how often Google shows your links in AI Overviews and AI Mode, how often AI assistants name or cite your pages, and how many visits those assistants send you. Google Search Console shows the first one as impressions only. Google Analytics 4 shows the third one as AI Assistant visits. The second one, mentions and citations in ChatGPT, Gemini, Perplexity and Claude, has no Google report, so you test it yourself with a fixed set of prompts, repeated on a fixed schedule.

What AI visibility means

AI visibility has two layers, and they are not the same.

A mention is when an AI answer names your brand in its text, with no link to your site. A citation is when the answer lists one of your pages as a source. A linked citation is when that citation is also a clickable link to your site.

A mention can build awareness. A citation shows which of your pages the AI system trusts for that topic. A linked citation is the only one of the three that can send a visit directly. Record all three separately. Do not add them into one total and call it visibility.

A recommendation is a fourth layer to note when you see it: the answer does not only name you, it suggests you for the task. Record that too, but keep it separate from mentions and citations.

This article is about measuring. For how to rank, not just measure, see our guide on AI Search Visibility.

What Google actually gives you: the Generative AI performance report

Google Search Console has a dedicated report for this, called the Generative AI performance report. Google announced it on 3 June 2026, in a post by Hillel Maoz and Moshe Samet of the Search Console team, on the Google Search Central Blog: “Introducing Search Generative AI performance reports in Search Console“.

The report covers Search, for AI Overviews and AI Mode together, and there is a separate report for generative AI features in Discover. This article is about the Search report.

What it shows, in Google’s own words in that announcement: impressions, meaning how often URLs from your site appeared in generative AI features in Search and Discover; pages, meaning which URLs appeared within AI features; countries; devices, for Search results; and dates, with hourly, daily, weekly and monthly granularity.

Google says in Search Console Help that the reports reached all websites worldwide as of 31 August 2026.

What it does not show is just as important:

• No clicks. The report has impressions only.
• No queries. You can see that a page was shown in AI features, not what was asked.
• No split between AI Overviews and AI Mode. Both are counted together.
• No position and no CTR in this report.
• No API. The report is in the Search Console interface and export only, not in the Search Analytics API.
• Search Labs experiments are excluded, and Google says it may add more features to the list over time.

The same impressions are also counted in the main Performance report. The new report does not add new traffic. It isolates the AI slice for the first time.

I checked ClickRank.ai this way in the week of 18 to 24 October 2026. Clicks were 2,730, flat. Impressions were 779K, up from 730K. CTR was 0.4%, flat. Average position was 16.4, down from 14.8.

How to open it, step by step

1. Open your property in Google Search Console.
2. Go to Performance, then open the Generative AI performance report for Search. The direct address ends in /performance/search-analytics/ai.
3. Set the date range. Start with the last 28 days, then compare it with the previous 28 days.
4. Look at the Pages tab first. Sort by impressions. These are the pages Google is showing in AI Overviews and AI Mode.
5. Then check Countries and Devices. A page can be strong in one country and invisible in another, and that is a real finding, not a reporting error.
6. Export the chart and the table. The usual table limit of 1,000 rows applies. Keep the export, so you can compare the same pages next month.

If the report is not there, or it is empty, that is itself a signal. It usually means Google has not recorded enough AI impressions for your site yet, not that something is broken. Check again after a month, and check the main Performance report in the meantime.

What clicks mean, and where they live

Clicks from AI Overviews and AI Mode are in the main Performance report, under the Web search type, mixed in with all your other web clicks. Google says this in its guide “AI features and your website“: sites appearing in AI features are included in the overall search traffic in Search Console, reported in the Performance report within the “Web” search type.

We covered the November 2026 Core Update separately, including what to do during the roll out, when clicks and AI impressions can move in different directions.

There is no filter in the Performance report that separates AI-feature clicks from classic clicks. Do not report a click number as “AI clicks” unless you say clearly that it is a mixed total.

Two counting rules from Google are worth knowing when you read that report:

• A click on a link in an AI Overview counts as a click, in the same way as a click on a classic result.
• An AI Overview occupies a single position, and all its links share that position.

That is why a page can gain AI impressions in the Generative AI report while its clicks in the Performance report stay flat. The impression means the link was shown. It does not mean the person clicked it. Google also says, in the same AI features guide, that when people do click from search results pages with AI Overviews, those clicks are higher quality, meaning users are more likely to spend more time on the site.

On ClickRank.ai, in the week of 18 to 24 October 2026, AI Overview citations were 785+ pages and AI Mode citations were 1,025+ pages, both up 15, while clicks stayed flat at 2,730. For “deepseek vs chatgpt”, position was 4.05, impressions 148,267, clicks 15. For the brand query “clickrank”, CTR fell from 42.8% to 15.8% while position improved to 1.47.

GA4: the AI Assistant channel

Google Analytics 4 has a built-in channel for visits from AI assistants. Google added it on 13 May 2026, in its Analytics release notes, as “New AI Assistant traffic measurement”.

When a session’s referrer matches a recognised AI assistant, GA4 sets the medium to ai-assistant, the campaign to (ai-assistant), and groups the session under the AI Assistant channel in the Default Channel Group. Google’s default channel group definition describes it as traffic from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok.

Three limits to record with the number:

• It is referrer-based. A visit with no referrer cannot be recognised, whatever the assistant.
• It is not retroactive. Visits before 13 May 2026 were not re-labelled.
• It excludes Google’s AI Overviews and AI Mode. Clicks from those features stay in Organic Search in GA4, because they came from Google Search, not from an assistant app.

How to read it, step by step

1. In GA4, open Reports, then Acquisition, then Traffic acquisition.
2. Keep the primary dimension on Session default channel group and look for the row called AI Assistant.
3. Then switch the dimension to Session source and search for chatgpt, perplexity, gemini, copilot and claude. That shows which assistant sent each visit, including visits that landed in Referral or Unassigned instead of the new channel.
4. If you want your own list, add a custom channel group with a source regex, placed above Referral. A custom group applies to past data too, unlike the built-in channel.

Read the number as a minimum, not a total. Mentions with no click, assistant fetches, and visits with stripped referrers will never show up in GA4.

Bing Webmaster Tools: the AI Performance report

Bing Webmaster Tools has its own AI Performance report, for Microsoft Copilot and Bing’s AI answers. It reports citations, the pages cited, and the grounding queries behind them. It does not report clicks from those answers.

Use it as a second check, not as a replacement for Search Console. If Bing cites a page that Google does not show in AI features, that page is worth a closer look: the content may be clear enough for one AI system and not yet clear enough for the other. Do not add Bing citations and Google impressions into one total. They count different things.

For Google AI Overviews specifically, ClickRank’s AI Overview Rank Tracker detects when an AI Overview appears for your keywords, extracts the cited sources, and compares sessions when an Overview is present versus absent.

The prompt test: the part Google cannot give you

ChatGPT, Perplexity and Claude offer no report to site owners on how often they cite your site. For them, you test it yourself. This is the part most AI visibility reports skip, and it is the part that tells you who is winning on the same question.

If a page is not indexed for ChatGPT at all, start with how to get indexed in ChatGPT Search before you run the test.

Build a fixed prompt set

1. Pick 20 to 30 prompts that reflect real buyer questions, not your own brand name alone. Include the questions people ask before they buy, and the comparison and how-to questions in your category.
2. Cover the engines your buyers use: ChatGPT, Gemini, Perplexity and Google AI Overviews and AI Mode, plus Claude and Copilot if your buyers use them.
3. Fix the account, location and settings you use for the test, and use the same ones every time.
4. Ask each prompt three times, on the same day. AI answers vary between runs, so one run is not a result. A trend across runs is.
5. Log every answer in a spreadsheet, one row per answer, including the answers where you do not appear. Do not log only the good ones.
6. Repeat the whole test monthly, in the same week of the month, with the same prompts.

What to record for each answer

• Mention: did the answer name your brand? Yes or no, and where in the answer it appeared.
• Citation: did it list one of your pages as a source? Which URL, and was your brand also named?
• Linked citation: did that citation include a clickable link to your site?
• Competitors: who was named or cited instead of you, on the exact same prompt?
• Accuracy: did the answer describe your product or service correctly? Note any error, because a wrong mention is not a win.

How to measure AI visibility: the five metrics to report each month

Report these five together, every month, so the picture cannot be read from one number alone:

1. AI impressions (Search Console): total impressions in the Generative AI performance report, and the top 10 pages by impressions. No clicks, no queries, by Google’s design.
2. Citation rate: the share of your test prompts where at least one of your pages was cited. Divide cited prompts by total prompts tested.
3. Mention rate: the share of your test prompts where your brand was named, with or without a citation.
4. Linked citation rate: when you were cited, how often did the citation include a link to your site?
5. AI referral sessions (GA4): sessions in the AI Assistant channel, by source, plus the Organic Search sessions for the same period, kept separate. Do not mix them.

Add a sixth line if you use Bing: Bing AI citations, by page, kept separate from the Google numbers.

For ClickRank.ai, my baseline in the week of 18 to 24 October 2026 was 975+ mentions across ChatGPT, Perplexity and Gemini, up 15. I record that baseline next to the citation rate, not instead of it.

Share of voice is the comparison that makes these numbers useful: your presence, against your competitors, on the same prompts, in the same month. Record which domains get cited for each prompt, yours and everyone else’s, then compare the shares. A rise in your citation rate means little if a competitor rose faster on the same prompts. For the difference between the three approaches behind those citations, see AEO vs GEO vs SEO.

What not to do

• Do not report CTR as AI visibility. CTR tells you nothing useful when the answer lives inside ChatGPT or an AI Overview. A page can influence thousands of AI answers and register zero clicks.
• Do not report one prompt run as a result. Answers change between runs. Count how often you appear, not where you appeared in one answer.
• Do not report GSC AI impressions as clicks. Impressions are how often a link was shown. Clicks are in the main Performance report, mixed with everything else.
• Do not claim ChatGPT, Perplexity or Claude have a Search Console-style report. They do not, at the time of writing. For them, you rely on referral traffic in GA4 and on your own prompt test.
• Do not buy a tool on the promise of one number. Ask any vendor which engines it measures, whether it uses the consumer app or an API, and whether it counts mentions, citations or linked citations. API answers do not behave exactly like the consumer app.

If visibility is low, what to fix first

Google says, in its AI features guide, that there are no additional requirements to appear in AI Overviews or AI Mode, and no other special optimizations necessary. The one technical condition is that a page must be indexed and eligible to be shown in Google Search with a snippet.

Start there, then work through this order:

1. Check the page is indexed, and that crawling is allowed in robots.txt. Use the URL Inspection tool in Search Console to see the HTML Googlebot received.
2. Make the answer easy to find in the text. Put the direct answer at the top, in short paragraphs, with one clear point each. Make sure important content is available in textual form, not only in an image or a video.
3. Add real expertise and original value: original data, a case study, first-hand experience, or a specific example. If you cannot name at least 2 or 3 genuinely unique elements on a page compared with its top 3 competitors, that page needs substantive improvement, not a light rewrite.
4. Check E-E-A-T signals: a named author with real credentials, an author bio with relevant experience, and links to primary sources. Do not invent credentials. For more on this, see our guide on E-E-A-T.
5. Keep the content fresh and dated. For more on this, see our guide on content freshness.
6. If a page is not appearing in AI Overviews at all, check the guide on why content does not appear in Google AI Overviews.
7. For the wider method, see our ChatGPT SEO guide and our guide on AI Engine Optimization (AEO/GEO), and our Google Search Console guide.

Do not rewrite a whole site because one month’s AI impressions fell. Wait for the next monthly test, compare the same prompts, and change one thing at a time, so you can tell what helped.

Experienced Content Writer with 15 years of expertise in creating engaging, SEO-optimized content across various industries. Skilled in crafting compelling articles, blog posts, web copy, and marketing materials that drive traffic and enhance brand visibility.