Measuring AI search visibility in 2026: Search Console's generative AI report, Bing's AI Performance and GA4

As of October 2026 you can measure a good part of your AI search visibility with free, first-party data: Search Console’s generative AI performance report counts impressions in AI Overviews and AI Mode, Bing Webmaster Tools counts how often Copilot and Bing’s AI summaries cite you, and GA4 has had a default “AI Assistant” channel for visits from ChatGPT, Gemini and others since May 2026. What none of them shows: clicks from Google’s AI features as a separate number, the queries behind your AI Overview impressions, or how often ChatGPT, Claude or Perplexity mention you without anyone clicking. A useful AI visibility report is built from these three sources and says plainly what they leave out.

Most of what ranks for this topic is written by companies selling prompt trackers, and their conclusion is predictably that you need one. Guides written before May 2026 also tell you to build a GA4 regex channel for AI traffic from scratch; GA4 now does a version of this by default. Below is what each source actually measures, according to its own documentation, as of October 2026.

What you can measure now, and what you still can’t

Question	Source	Available?
How often were links to my site shown in AI Overviews and AI Mode?	Search Console, generative AI performance report	Yes, impressions by page, country, date, device
How many clicks came from AI Overviews or AI Mode?	Search Console, GA4	No. Included in total organic clicks, not separable
Which queries triggered my AI Overview impressions?	Search Console	No query dimension in the generative AI report
How often do Copilot and Bing’s AI summaries cite my pages?	Bing Webmaster Tools, AI Performance	Yes, citations per page and sampled grounding queries
How many visits come from ChatGPT, Gemini, Copilot and co.?	GA4, AI Assistant channel	Yes, if the visit arrives with a referrer GA4 recognises
How often does ChatGPT, Claude or Perplexity mention me without a click?	Nobody’s first-party data	No. Only prompt sampling gets near it

The pattern: Google tells you about impressions, Bing about citations, GA4 about visits. You need all three, and you can’t add them up.

Search Console: the generative AI report, read correctly

Google announced the generative AI performance reports on 3 June 2026, first for a subset of sites. Since 31 August 2026 they are available to all websites worldwide. There are two: one for Search and one for Discover.

The Search report covers AI Overviews and AI Mode, not Search Labs experiments. Its one metric is impressions: “how many times links to your site were shown to a user in a generative AI feature on Google Search.” You can group them by page, country, date and device, and filter by search type: text queries or multimodal searches such as Google Lens and Circle to Search.

Four details change how you read it:

• No clicks, no CTR, no position, no queries. The report has impressions and four dimensions. Google’s announcement says it may add “additional metrics over time”, without a date.
• The chart counts per site, the page table per page. If two of your pages appear in the same AI response, the chart counts one impression; the page table counts both. That is why chart and table totals differ. Add a URL filter and the chart switches to URL-level counting.
• It is a subset of data you already have. The report “includes data from the Web search type in the Performance report”. Your AI Overview impressions and clicks were always in the normal Performance report; the new report only isolates the impressions.
• The usual Performance report limits apply, including the 1,000-row table limit. The export button downloads chart and table; values shown as ~ or - become zeros in the file. The help page documents the export but says nothing about the Search Console API.

If the report doesn’t appear at all, Google names two reasons: too few impressions in these features, or your site is excluded from them via the Search generative AI setting. That setting and what excluding costs are covered in my article on AI crawler controls.

The Discover report works the same way with one difference worth knowing: it has pages, countries and dates but no device dimension, and it counts each page separately even when two of your pages appear in the same Discover list.

Why your average position looks odd

The normal Performance report keeps the clicks. Google’s methodology page says a click on an external link in an AI Overview or in AI Mode counts as a click, and that “an AI Overview occupies a single position in search results, and all links in the AI Overview are assigned that same position.” In AI Mode, a follow-up question counts as a new query. So a page cited in an AI Overview at the top of the results can report position 1 with very few clicks. Before you call that a CTR problem, check whether the page has AI impressions.

The one ratio worth tracking

Because the generative AI data is a subset of the Web search data, you can put the two side by side per page and period:

1. Export the generative AI report by page for the last 28 days.
2. Export the normal Performance report (search type Web) by page for the same 28 days, with clicks and impressions.
3. Join on URL and compute AI impression share = AI impressions ÷ total Web impressions.

Pages with a high AI impression share and falling clicks are the ones to look at first: Google uses them in AI answers, and fewer people continue to the page. That fits independent research. In Pew Research Center’s analysis of the browsing data of 900 US adults in March 2025, users clicked a traditional result in 8% of visits when an AI summary appeared and in 15% when none did; they clicked a link inside the summary in 1% of visits. That is US data from 2025, not a forecast for your site, but it explains why impressions and clicks drift apart.

Track the ratio by page group (product pages, guides, category pages), not by single URL. Single pages swing too much from week to week to read anything into.

If a page you’d expect in AI answers has zero AI impressions, check the basics before anything else. My Index Check tests a URL for the technical blockers that keep it out of the index: status code, noindex, robots.txt, canonical and sitemap. A page that isn’t indexed can’t be shown in AI Overviews either.

Bing Webmaster Tools AI Performance: citations and grounding queries

Microsoft launched AI Performance as a public preview on 10 February 2026. It covers Microsoft Copilot, AI-generated summaries in Bing and “select partner integrations”. Its metrics:

• Total citations: how often your content was shown as a source in AI answers in the period.
• Average cited pages: unique pages from your site cited per day.
• Page-level citations and a trend over time.
• Grounding queries: the phrases the AI used when it retrieved your content. Microsoft notes that the data “represents a sample of overall citation activity”. These are retrieval queries, not the prompts users typed.

On 16 June 2026 Microsoft added four preview features: Intents (grounding queries grouped into categories such as informational, commercial or navigational), Topics (thematic clusters), Citation Share (your share of all citations shown for a grounding query) and Compare (overlay a previous period). Microsoft is explicit that Citation Share is “an observational metric – not a ranking system or a competitive scoreboard”; it doesn’t show competitor domains or traffic share.

Neither announcement mentions clicks. What Bing gives you that Google doesn’t is a query-like signal: grounding queries show which topics an AI system pulled your pages for. I use them as a content input, to find topics where you are cited for one sub-question but not the obvious next one. If Bing sends you little traffic, read the citation trend, not the absolute numbers.

GA4: the AI Assistant channel, and when to build your own

On 13 May 2026 Google Analytics added an “AI Assistant” channel to the default channel group. According to the default channel group documentation:

• The channel covers visits “from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok”.
• It “excludes Google’s AI Overviews and AI Mode”. Those clicks are in Organic Search.
• Mechanically, GA4 sets the medium to ai-assistant and the campaign to (ai-assistant) “if the referrer matches a list of AI Assistants”. The help page doesn’t say which referrers are on it.

Three consequences follow. First, clicks from Google’s AI features are invisible as a separate number in GA4, just as in Search Console. Second, the channel only works when a referrer arrives. A visit from an app or browser that sends no referrer lands in Direct, and no channel rule can recover it. Third, because the list isn’t public, check where your AI sources actually land: in Reports > Acquisition > Traffic acquisition, switch the primary dimension to Session source / medium and search for the assistant domains you care about. If one of them shows up as referral instead of ai-assistant, the default list doesn’t cover it.

That is the case for a custom channel group. Google’s custom channel group help page includes an “AI assistants” example and says to place the channel above Referral, because traffic goes to the first channel whose rule matches. A version I find easier to maintain matches on Source:

.*(chatgpt|openai|perplexity|claude|gemini|copilot|deepseek|grok).*

Two rules from Google’s regex documentation matter here: GA4 regex is a full match by default, which is why the pattern is wrapped in .*, and it is case sensitive. Before you save, compare the pattern against your actual source list; gemini should match gemini.google.com, and you don’t want it to catch anything else. Standard properties allow two custom channel groups, and they apply to your reports retroactively.

For leads, the channel matters more than the sessions. Add key events (form submissions, demo requests, purchases) to the Traffic acquisition view by channel. A small channel with a high conversion rate is a different conversation with management than a small channel full of bounces.

Prompt tracking tools: when they’re worth paying for

Prompt trackers send a fixed list of prompts to ChatGPT, Perplexity, Gemini and others on a schedule and record whether your brand or URL appears. They cover the one gap first-party data can’t: mentions without a click in assistants that don’t report to you.

Their limits are built into the method. AI answers are generated per request and can differ from one run to the next, so a tracker samples; it doesn’t count. The prompt list is written by you or the vendor, not taken from real user logs. And a “visibility score” combines these samples in a way each vendor defines differently.

My rule for whether to pay:

• Not yet, if the AI Assistant channel in GA4 and the generative AI report in Search Console show little activity. Fix crawlability, indexing and content first; a tracker will only confirm the absence.
• Worth it, if AI assistants already bring measurable leads or revenue, you compete on a set of commercial questions you can name (“best X for Y”, “X vs Y”), and someone will act on the results monthly.
• Either way, keep the prompt list short, commercial and stable, and read trends over months, not single runs.

A one-page AI visibility report for management

This is the structure I’d use. Each line says what it measures and what it doesn’t.

KPI	Source	What it says	What it doesn’t say
AI Overview + AI Mode impressions (monthly, by page group)	Search Console generative AI report	How often Google showed links to you in AI features	Clicks, queries, position in the answer
AI impression share by page group	Generative AI report ÷ Performance report (Web)	Which content Google uses in AI answers	Whether that helps or costs you traffic
Organic clicks and CTR on high-AI-share pages	Search Console Performance report	Whether those pages still get visits	Which clicks came from the AI feature
Bing / Copilot citations and top grounding topics	Bing Webmaster Tools AI Performance	Where Microsoft’s AI cites you	Clicks; the user’s actual prompt
AI Assistant sessions and key events	GA4 AI Assistant channel (plus custom group)	Visits and conversions from ChatGPT, Gemini, Copilot and others	Visits without referrer; Google AI clicks
Mention rate on 20–50 commercial prompts (optional)	Prompt tracker	Presence in answers without clicks	Real query volume; anything beyond the sample

Put one sentence under the table every month: what changed, why you think it changed, and what you’ll do about it. Don’t add the rows into a single “AI visibility score”. Impressions, citations and sessions measure different things in different systems.

Where to start this week

1. Open the generative AI report in Search Console and export it by page.
2. Join it with the Performance report and find the ten page groups with the highest AI impression share.
3. In GA4, check Session source / medium for AI assistant domains and confirm they land in the AI Assistant channel.
4. Verify your site in Bing Webmaster Tools if you haven’t, and read the grounding queries for your most cited pages.

If AI visibility is new territory for your team, my AI search checklist goes through crawler access, rendering, quotable content, entities and measurement. If you’d like the setup and the first report built for your site, that’s part of my AI search work.