Check this page with an assistantOpens a chat asking it to summarise this article and name the evidence behind each claim.
Claude opens with the prompt on your clipboard: Anthropic does not support prefilled prompts on the web, and we would rather copy it than ship a button that drops it.
What Actually Shipped on 3 June 2026
Google announced dedicated Search generative AI performance reports for Search and Discover. They show, in Google's wording, how often URLs from your site appeared in generative AI features in Search and Discover. The dimensions available are impressions, pages, countries, devices, and dates. Clicks, click-through rate, and query data are absent.
Two things about that announcement matter more than the feature itself. The first is that this is a breakout, not new data. Google was explicit that generative AI impressions were already inside the overall performance report and will continue to be tracked there. Your headline numbers did not change on 3 June; a slice of them simply became visible. Anyone telling you their traffic moved because of this report is describing a coincidence.
The second is the rollout. At announcement the reports were going to a subset of website owners in the UK, with global expansion planned and no date attached. So for most properties reading this, the correct action today is not to open Search Console and hunt for a menu item. It is to understand what the instrument measures before it arrives, so that the first thing you do with it is not overread it.
You can state, in one sentence, the difference between an impression in this report and a visit to your site.
The Four Questions It Can Genuinely Answer
Impressions-only data still answers real questions, as long as every question is about eligibility and selection rather than value. Which pages AI features are willing to surface, whether that set is growing, whether it matches the pages you care about, and whether a change you shipped moved it.
- Which pages are AI-visible at all?
Sort by impressions and read the top twenty URLs. This is the set of pages Google's systems consider quotable on some query. If your highest-margin service page is absent while an old blog post dominates, you have found a genuine gap, and it is usually a structure problem rather than a topic problem.
- Is the AI-visible set growing or narrowing?
Track the count of distinct URLs with non-zero impressions month over month, not just total impressions. Total impressions can rise because one page got lucky. Distinct URLs rising means more of your site became eligible, which is the thing you can actually influence.
- Does AI visibility match commercial priority?
Put your ten most valuable pages in a column and their AI impressions beside them. The mismatch is the brief. Pages that convert but never surface are the highest-leverage rewrite candidates on the site, and they are invisible in every other report you run.
- Did a specific change move anything?
Because the date dimension goes down to fine granularity, you can put a change date against a page-level impression line. This is the closest thing to a controlled test AI visibility has offered so far. It is still observational, so treat a single page moving as a hint and a template-wide move as evidence.
Notice that all four are questions about coverage. That is the honest ceiling of an impressions-only dataset, and it is a useful ceiling. Coverage is the gate that everything else waits behind: a page that never surfaces cannot be cited, clicked, or credited.
What an Impression Cannot Tell You
An impression in this report means a URL from your site appeared inside a generated answer. It does not mean a person read your name, noticed the citation, agreed with the summary, or came to your site. Impressions establish that you were retrievable and selected. Everything after that is unmeasured here.
This is where the reporting discipline has to hold. The temptation, especially if someone is paying for AI visibility work, is to convert impressions into a number that sounds like performance. It cannot be done honestly with this dataset. There is no click column, so there is no rate to compute; there is no query column, so you cannot say which questions you won; there is no competitor column, so you cannot say whether you were named first, third, or in passing.
| Question | Can this report answer it? | Where the answer actually lives |
|---|---|---|
| Did my page surface in an AI answer? | Yes | This report, page dimension |
| Is AI visibility trending up? | Yes, for impressions and distinct URLs | This report, date dimension |
| How many visits did AI features send? | No | Overall performance report, plus a corrected GA4 channel setup |
| Which questions did I get cited on? | No | Manual prompt sampling; see share of model |
| Am I named ahead of competitors? | No | Prompt sampling, logged over time |
| Did AI-referred visitors convert? | No | Analytics and CRM, with source preserved |
The right posture is the one this report itself takes: state what was measured, name what was not, and refuse to fill the gap with an estimate. That is the same rule we apply to the Growth Map, where scoring components that need Search Console or analytics data are declared missing rather than guessed.
Where the Click Data Has Been All Along
AI Mode clicks and impressions have been folded into the standard Search Console performance report since June 2025. A click on an external link inside AI Mode counts as a click, an appearance counts as an impression, and position follows the same methodology as an ordinary results page. What you cannot do is filter to see only that traffic.
This is the part most write-ups get backwards. The June 2026 launch did not begin counting AI traffic; it began separating a view of AI impressions. The clicks were already in your totals, silently, mixed with everything else. If your organic clicks held steady through 2025 and 2026 while impressions climbed, some of that divergence is AI surfaces generating appearances that resolve without a visit, which is the mechanic behind zero-click searches.
Two practical consequences. First, do not go looking for a hidden pool of AI clicks to add to your reporting; you would be counting them twice. Second, the useful comparison is between the new impressions view and your existing performance report on the same pages and dates. Where AI impressions rise and total clicks for that page do not, you are watching an answer get resolved without you. The fix is rarely more content; it is usually giving the page something a summary cannot carry, which is the subject of our guide to what to do when an assistant sends you a visitor.
The Opt-Out Toggle, and Who It Is Actually For
Google shipped a control letting sites block their content from appearing in generative AI features, and stated that sites which opt out will not receive traffic or impressions from those features, while the choice will not be used as a ranking signal for search results outside them. Regular Search rankings are therefore not the cost. The forfeited referrals are.
Read that as a straightforward trade rather than a moral question. You are choosing between your content being summarised in answers that may not send a visit, and your content being absent from a surface where buyers increasingly start. For most small and mid-sized sites the second is worse, because the visits that do come through tend to arrive with the category already understood and a decision partly made.
There are real cases on the other side. Subscription publishers whose entire product is the text. Sites under a licensing arrangement that makes free summarisation a contract problem. Businesses whose content is so easily condensed that the summary genuinely substitutes for the visit. If you are none of those, the toggle is not your lever, and the related question of which crawlers to allow is a separate and more useful decision, covered in our decision guide for AI crawler access.
Export twelve months of performance data first. Opting out changes what future data can exist, and there is no version of this decision you can evaluate without a baseline.
Adding It to the Weekly Review Without Adding an Hour
Three metrics, once a month, in the review you already run: total AI impressions, count of distinct URLs with impressions, and the overlap between the top twenty AI-visible URLs and your top twenty commercial pages. Everything else in the report is context for those three.
The reason to cap it at three is that this data invites drifting. There is a country dimension, a device dimension, and hourly granularity, none of which will change a decision for a small site, and all of which will happily consume a morning. The weekly routine that does change decisions is the striking-distance and decay work in the standard performance report, which is unchanged and still where the leverage is; that routine is laid out in the Search Console operator's guide.
- Log the three numbers.
One row per month in the same sheet as the rest of your search reporting. The value is the trend line, and a trend line needs a boring, consistent record more than it needs precision.
- Diff the AI-visible URL set.
Which URLs entered the set, which left. Departures are worth more attention than arrivals: a page that stopped being quotable usually changed, or its competition did.
- Write one sentence of interpretation.
Not a paragraph. If you cannot state what changed and what you intend to do, the month's answer is "no signal", and writing that down honestly is what keeps the record trustworthy when a real signal shows up.
The measurement is only half the loop. The other half is deciding which pages deserve the work, which is the job the Search Console opportunity finder does against your performance export, and the wider reporting frame it belongs to is in our SEO reporting framework.
Questions People Ask About the AI Performance Report
- What is the Search Console generative AI performance report?
It is a dedicated view, announced by Google on 3 June 2026, showing how often URLs from your site appeared inside generative AI features in Search and Discover, including AI Overviews and AI Mode. It reports impressions broken down by page, country, device, and date. It does not currently report clicks, click-through rate, or query data.
- Does the new report change my overall Search Console numbers?
No. Google has said that generative AI impressions were already included in the overall performance report and will continue to be tracked there. The new reports break that existing data out into a dedicated view rather than adding traffic you were not already counted for. If your totals moved, something else caused it.
- Why does the report show impressions but not clicks?
Google described the launch as a first step and said it is continuing to work with website owners to understand which insights are most helpful, with additional metrics to follow. Until clicks arrive, an impression in this report tells you that a URL surfaced inside an AI answer, not that the surfacing sent anyone to your site or was worth anything.
- Should I use the toggle to block my content from AI features?
Only if you can afford to lose the traffic. Google has stated that sites opting out will not receive traffic or impressions from its generative AI features, and that the choice will not be used as a ranking signal for search results outside those features. So regular Search rankings are not the risk; the forfeited AI referrals are. For most small sites that trade is a bad one, because AI-referred visitors tend to arrive further along in a decision.
- Why can I not see the report in my Search Console?
At announcement the reports were rolling out to a subset of website owners in the UK, with global expansion planned but undated. If your property is outside that group the section will not appear yet. There is nothing to configure and no way to request access; the data is being generated regardless of whether the view is visible to you.
Primary Sources
- Google Search Central: Introducing Search generative AI performance reports in Search Console (3 June 2026)
- Google Search Central: AI features and your website
- Google Search Console: Performance report reference
- Search Engine Land: Search Console AI performance reports and controls to block content in AI responses
- Search Engine Land: AI Mode traffic data in Search Console
