GSC’s new Generative AI report doesn’t track CTR—here is your new KPI for AI search visibility

On 3 June 2026, Google put a number on something we’d only been able to estimate: how often your pages show up inside its AI answers. The new Generative AI report in Search Console tracks impressions in AI Overviews and AI Mode—and pointedly omits click-through rate. That omission is the whole story.

The week AI search became measurable

Three things happened inside ten days, and they’re connected. Google’s May 2026 broad core update started rolling out on 21 May and finished on 2 June, with the usual stated intent to better surface relevant, satisfying content for searchers from all types of sites. One day later, on 3 June, the Generative AI performance report appeared in Search Console—rolling out to a subset of UK site owners first, with global expansion promised but no date attached. And all of this lands while AI Overviews have become a fixture of everyday search results rather than an experiment.

For the first time, B2B site owners have a Google-native instrument for AI impression data. That matters because every third-party tool until now was inferring AI visibility from the outside. This report reads it from the inside. But here’s the catch I want you to sit with before you open the report: it’s built on a fundamentally different logic than anything in the Performance Report you already know. Treat it like a second clicks-and-CTR dashboard and you’ll misread it completely.

What the report actually shows

You’ll find it in Search Console under Performance, as a separate Generative AI entry alongside the standard Search results and Discover views. It ships in two flavors—a Search-specific view and a Discover-specific view. Open it and you get impressions segmented four ways: by page, by country, by device, and by date, at a granularity that runs from hourly all the way to monthly.

The definition of impression here is the part most people will skim past, and it’s the part that changes how you read everything. Per Google’s own documentation, impressions in this report are how many times links to your site were shown to a user inside a generative AI feature on Search. Not how many times your brand was named in prose—how many times a clickable link to your page was rendered in an AI answer.

There’s a counting nuance worth internalizing. A standard blue-link impression registers the moment the results page loads, regardless of where the link sits or whether anyone scrolls to it. An AI Overview link is counted differently: it has to be scrolled or expanded into view to register. So an AI impression is a slightly higher-quality signal than a raw SERP impression—it means the citation slot was actually surfaced to a human eye, not just buried below the fold.

What it deliberately omits

No clicks. No CTR. No average position. No query strings. Google has acknowledged the gap and signaled that more is coming, but for now the report tracks impressions, pages, countries, devices, and dates—and stops there.

Read that as a design decision, not a privacy workaround. The click funnel model—impression, click, session, conversion—simply doesn’t map onto a generative answer surface. A buyer can read your data, your framing, and your name inside an AI Overview, form a view of your company, and move on to the next step in their research without ever clicking through to you. The influence happened. The click didn’t. CTR can’t measure something that, by design, doesn’t produce a click.

The numbers backing this up have only gotten starker. Ahrefs’s April 2025 study of 300,000 keywords found that the presence of an AI Overview correlated with a 34.5% lower average CTR for the top-ranking page—position-one CTR fell from 0.073 in March 2024 to 0.026 a year later. When they re-ran the study on December 2025 data and published the update in February 2026, the gap had widened to 58%. The click-centric KPI era isn’t ending gradually. It’s already most of the way gone, and this report is Google quietly admitting it.

Your new primary KPI: AI Impression Share

If you can’t measure clicks, measure presence. The metric I’d put at the center of any AI search dashboard is AI Impression Share: for each page, the ratio of its AI impressions to its total search impressions. That single ratio tells you how much of a page’s visibility is now happening inside answer surfaces rather than the blue-link list—and it normalizes across pages of wildly different traffic.

Three ways to make it useful:

• Week-on-week trend as a proxy for citation consistency. A page that holds or grows its AI Impression Share is one Google trusts repeatedly to ground an answer. A jagged, collapsing line is a page being cited by accident, not by reputation.
• Cross-reference against the standard Performance Report. Match your high-AI-impression pages against your high-click pages. Where they overlap, you’ve found content that both earns the citation and converts the click—your crown jewels. Where a page has heavy AI impressions but thin organic clicks, that’s a page doing invisible work you were never crediting.
• Overlay the May Core Update. Line up which pages gained and lost in the core update against which pages now carry AI impressions. Strong thematic coherence between the two is a signal that Google reads that page as a trustworthy source on its topic—exactly the proximity the update rewarded.

In every audit I run, the first forensic question is which pages Google trusts enough to surface in a zero-click answer slot. Until this month I had to triangulate that. Now we can track it directly.

The AI-blocking control—and when to actually use it

The report ships next to a control I expect a lot of people to misuse: an opt-out toggle that excludes your content from AI Overviews and AI Mode without touching your organic rankings. Google begins enforcing the setting on 17 June 2026. The trade-off is total: opt out and you lose all impressions—and any traffic—from those generative surfaces. You also go dark in this new report, because there’s nothing left to count.

There are narrow cases where opting out is the right call:

• Paywalled content whose entire value depends on the click. If the AI answer satisfies the reader, you’ve given away the product.
• Proprietary methodology pages where a paraphrased citation strips the context that makes the work defensible—and misrepresents it in the process.
• Highly regulated content—financial, medical, legal—where an AI rewording of your wording creates liability you can’t control.

And there’s the case where it’s almost always wrong: any page a B2B buyer might hit while researching a vendor in Gemini, ChatGPT, or AI Mode. If you sell to businesses, opting your comparison pages, case studies, and product explainers out of AI answers is volunteering to be absent from the exact moment a decision is forming. For most small and midsize companies the default is simple: stay indexed, measure for a few weeks, then decide page by page with data in hand.

Why DACH sites need to move faster than US peers

German-speaking B2B has a sharper version of this problem, for three reasons.

First, the click loss is already severe here. SISTRIX’s March 2026 analysis of more than 100 million German keywords found AI Overviews cut the click-through rate on Germany’s top organic position from 27% to 11%—a 59% drop—with an estimated 265 million clicks per month disappearing from the German organic market. That’s not a forecast. That’s last quarter.

Second, most DACH sites I audit aren’t blocked from AI answers—they’re simply never cited in the first place. No FAQPage, no Article schema, no datePublished. When the structured signals an AI system uses to ground and date a claim aren’t there, your pages don’t show up in the impression data because they were never in the running. Zero AI impressions can mean you opted out. More often it means you were never invited.

Third, the May Core Update rewarded what I’d call source proximity—first-party data, primary expertise, the page that is closest to the original fact rather than a summary of summaries. That maps almost one-to-one onto the GEO work we do: make the page the primary source, mark it up so machines can read it, and earn the citation on merit. The sites that won in May are, disproportionately, the ones already showing up in the new report.

Building an AI visibility reporting dashboard

You don’t need a data team for this. A workable dashboard for a three-person marketing function has three inputs:

1. GSC Generative AI impressions—the new report, pulled weekly, per page.
2. Standard Performance Report impressions and clicks—so you can compute AI Impression Share and spot the citation-plus-conversion overlap.
3. A weekly manual citation spot-check—run your 5–10 target queries through ChatGPT, Gemini, and Perplexity and record whether you’re named and how you’re framed. Google’s report tells you nothing about the non-Google engines, and this is the cheapest way to cover that blind spot.

The manual spot-check is fine at five queries and painful at fifty. Once you’re tracking more queries, or more competitors, than a person wants to re-run by hand each week, that’s the point to automate it—a tracker like Cited watches how ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews describe your brand and your competitors over time, so the spot-check becomes a trend line instead of a Friday chore. We built it because the manual version doesn’t scale past a handful of queries.

The one number for the executive dashboard isn’t a count of citations. It’s the AI Impression Share trend across your revenue pages. Up and to the right means Google is trusting you more inside answers. Flat or falling means your competitors are taking the slots—and you’d never have known from a clicks report.

Week-one action checklist

Ordered by effort against impact. Most of this is doable in an afternoon; two items are genuine projects.

1. Verify GSC access and find the report (5 minutes). Performance → Generative AI. If it’s not there yet, you’re outside the current rollout—check back as it expands globally.
2. Pull AI impression data for your top 10 revenue pages (30 minutes). Export it. This is your baseline; every later trend reads against it.
3. Cross-reference with the May Core Update (1 hour). Which pages lost ground in late May, and do they overlap with low or zero AI impressions? That overlap is your priority fix list.
4. Add datePublished and dateModified Article schema to every page with AI impressions (half a day). Dated, structured pages are easier for AI systems to ground a citation on. This is the single highest-leverage move on the list.
5. Confirm robots.txt allows the right bots (30 minutes, and read this twice). Training crawlers—GPTBot, ClaudeBot, Google-Extended—are not the same actors as the retrieval bots that fetch you for AI search answers (OAI-SearchBot, Claude-SearchBot, PerplexityBot). Roughly 71% of top news publishers block at least one retrieval bot while only meaning to block training. Make that mistake and you delete yourself from AI answers entirely.
6. Decide your AI-blocking posture page by page (1 hour) before the 17 June enforcement date. Default to staying indexed unless a page meets the narrow opt-out cases above.
7. Set a recurring four-week review cadence (15 minutes). AI Impression Share trend, the citation spot-check, and any movement against core-update winners and losers.

The bottom line

Here’s the claim, stated cleanly so you can quote it: GSC’s Generative AI report tracks impressions inside AI Overviews and AI Mode but strips out CTR entirely—confirming that citation share, not click rate, is the correct performance metric for AI search visibility in 2026.

Google didn’t forget to include clicks. It told you, by leaving them out, that you’ve been optimizing for the wrong number. The work now is to find out which of your pages Google trusts enough to surface in an answer, make more of them look like that, and measure the share—not the clicks.

If you’d rather see that mapped against your own domain than build the baseline yourself, that’s exactly what our AI Visibility Audit does: which pages get cited, which got left behind in May, and the schema and crawler fixes that move the AI Impression Share trend. Flat-rate, starting at 99 € net.