Since 31 August 2026 every Search Console property in the world has a report showing how often its pages appear inside Google's AI features. It is the first time Google has given site owners any direct measurement of AI visibility, and it is also the most widely misread report in the product, because almost everything people want from it — clicks, queries, positions — is not in it.
Search Console's generative AI performance report shows impressions only: how often your pages appeared in AI Overviews, AI Mode and Discover's AI features, broken down by page, country, device and date. There are no clicks, no queries and no average position. Use it to set an AI impression baseline now, then track the gap between AI impressions and Search clicks over time.
What the report actually contains
Google introduced the generative AI performance reports in Search Console in June 2026 and finished rolling them out to all properties worldwide on 31 August 2026. If you have not looked since the summer, it is there now regardless of your site's size or country.
An impression in this report means one of your URLs appeared inside a generative AI surface: cited in an AI Overview, referenced as a source in AI Mode, or surfaced in Discover's AI features. What you get, and what you do not, is unusually clear-cut:
| Dimension or metric | In the report? | What that means for you |
|---|---|---|
| Impressions | Yes | The only volume metric available. Your entire baseline has to be built on it |
| Pages | Yes | You can see which URLs Google's AI features pull from, which is the single most actionable field in the report |
| Countries | Yes | Useful if AI feature rollout differs by market, which it does |
| Dates | Yes | Lets you build a trend line, which is the point of starting now |
| Devices | Yes (Search report) | The Discover version of the report does not break out device |
| Clicks and CTR | No | You cannot tell from this report whether an AI citation sent you a visitor |
| Queries | No | You cannot see which questions triggered the citation, so keyword-level AI reporting is not possible here |
| Average position | No | There is no ranking equivalent inside an AI answer |
One detail catches people out: these impressions are not held separately from the rest of Search Console. They are also included in the overall Search performance report totals. So if your total impressions climbed this year while clicks did not, part of that gap has been AI feature impressions sitting inside your headline number the whole time. The new report is what lets you finally separate the two.
Why "impressions, no clicks" is the finding, not a limitation
The instinct is to treat the missing click data as an oversight. It is more useful to treat it as a description of the surface. Appearing in an AI answer is structurally different from ranking: the answer is designed to resolve the query in place, and the citation is a footnote on a page the user has already read.
The behavioural evidence is consistent on this point. Pew Research Center's study of real browsing behaviour found that users clicked a traditional search result 8% of the time when an AI Overview was present, against 15% when it was not. Ahrefs' analysis of AI Overview presence found the top-ranking page's click-through rate materially lower — in the region of 58% lower — on queries where an AI Overview appears.
Put those together with a report that counts impressions and cannot count clicks, and the strategic reading is straightforward: AI impressions are a brand and citation metric, not a traffic metric. Measuring them as though they were traffic is how teams end up reporting a healthy-looking AI presence next to a flat revenue line and no explanation for the gap.
Report AI impressions in their own row, never inside a traffic table. The moment they sit next to sessions in a dashboard, someone will divide one by the other and invent a CTR that does not exist.
The 20-minute AI visibility baseline
The report's data starts when your property got access, and Search Console's 16-month window will roll forward without you. Build the baseline now, because in six months the question "was this normal?" needs an answer you cannot reconstruct later.
Four numbers, in this order:
- AI impressions, last 28 days. From the generative AI performance report, unfiltered. This is your absolute AI visibility number.
- AI impression share. AI impressions divided by total Search impressions from the main performance report for the same 28 days. Express it as a percentage. This one number tells you how much of your visibility now lives in a surface that rarely sends a click.
- Cited page concentration. Export the pages table, then work out what share of AI impressions comes from your top five URLs. Concentration above roughly 60% means your AI presence rests on a handful of pages and is fragile; a long tail means the whole site is being read.
- The reference pair. Total Search clicks and total Search impressions for the same 28 days, recorded alongside the other three. Without this pair the first three numbers are uninterpretable next quarter.
Record all four in a sheet with the date, then repeat monthly. That is the entire method. It takes twenty minutes the first time and five minutes a month after that, and it is the difference between "AI traffic is down" as a feeling and as a measurement.
The four patterns, and what each one means
Once you have two or three months of that sheet, the movements start to mean something. Almost every real case falls into one of four patterns, and the correct response differs sharply between them.
| AI impressions | Search clicks | Most likely explanation | What to do |
|---|---|---|---|
| Rising | Falling | Working as designed: you are being cited, and the citation is answering the query in place | Shift the goal for those pages from traffic to citation. Add the specifics AI answers quote — dates, figures, conditions — and move the conversion ask to pages that serve later-stage queries |
| Rising | Rising | Your pages are cited on queries the AI answer cannot fully resolve, usually pricing, comparison or implementation | Your best position. Find which URLs do it and work out what those pages have in common, then apply it deliberately elsewhere |
| Falling | Stable | A competitor displaced you as a source, or Google changed which queries trigger AI features in your category | Check whether the affected pages still rank conventionally. If they do, this is a citation problem, not a ranking problem, and the fix is on the page: clearer direct answers, firmer sourcing, more specific claims |
| Falling | Falling | An ordinary visibility problem — ranking loss, indexing, a core update — not an AI-specific one | Diagnose it as a normal SEO issue. Do not let the AI framing send you down the wrong path |
The second row is the one worth chasing. A page that both gets cited and still earns the click is almost always doing something the AI answer cannot do in a paragraph: showing a real price range, comparing two options with conditions attached, or walking through a procedure with enough specificity that a summary is no substitute. That is a repeatable content instruction, and it is derived from your own data rather than from a general theory about AI search.
What the report cannot tell you, and the nearest workaround
Every serious question people bring to this report is out of its scope. Here is what each one actually requires.
| Question | In the report? | Nearest available answer |
|---|---|---|
| Did AI citations send me visitors? | No | Segment referrals in your analytics. AI assistant traffic arrives with its own referrers and can be grouped into a channel, then compared with the AI impression trend |
| Which queries cite me? | No | None inside Google's tools. Prompt-level testing against your own priority questions is the only substitute, and it is a sample, not a measurement |
| How do I compare with a competitor? | No | Not available. Search Console is first-party only; treat third-party "AI visibility share" figures as modelled estimates |
| Is ChatGPT, Copilot or Perplexity citing me? | No | Out of scope entirely. This report covers Google surfaces only, which is why referral segmentation stays necessary |
| Did an AI citation lead to revenue? | No | Only via last-click analytics on the referral channel, which will understate it. Assume attribution here is directionally useful at best |
The practical consequence is that the report is one of two measurements, not the whole thing. Pair it with a referral channel for AI assistants in your analytics and you have both halves: Google-side visibility from Search Console, and assistant-side traffic from referrals. We covered building the traffic half in landing pages for ChatGPT and Perplexity traffic.
The opt-out control: who should actually use it
Alongside the report, Google shipped a control that lets a site keep its content out of generative AI features in Search. Google states it is not used as a ranking signal for results outside those features.
For almost every commercial site, the answer is no. Removing yourself from AI features does not move your content back into the blue links a user would otherwise have clicked; it removes you from an answer that will be assembled from someone else's page instead. The narrow cases where it is worth considering are publishers whose business model is the page view itself, sites under a licensing arrangement that forbids this kind of reuse, and content with a legal or compliance reason not to be paraphrased. If you are a service business, an e-commerce store or a B2B site, opting out costs you citation visibility and returns nothing.
What to do in the next 30 days
- Record the baseline. The four numbers above, dated, in a sheet you will still have next year.
- Export the pages table. Sort by AI impressions and read the top 20 URLs. This is Google telling you which of your pages it considers quotable, which is information you have never had before.
- Cross-check those URLs against conversions. If the pages Google cites are not pages that lead anywhere commercially, that is the finding of the exercise, and it is a content-planning problem rather than a technical one.
- Fix the quotable pages first. On the URLs already earning AI impressions, tighten the direct answers, add dates and numbers to claims, and make sure each fact has a visible source. Pages that are already being read are the cheapest place to improve how you are represented.
- Set up the referral half. Group AI assistant referrers into their own channel in analytics so that next quarter's comparison is possible.
- Diary a monthly five-minute check. The value of this report is entirely in the trend, and the trend only exists if someone writes the numbers down.
None of this is an AI strategy in itself. It is the measurement layer that makes one possible, and it is worth doing before the next core update rather than after, because the argument you will be having in three months is about what changed — and that argument is unwinnable without a baseline.
Frequently asked questions
What does Search Console's generative AI performance report show?
It shows impressions from Google's generative AI features: AI Overviews, AI Mode and generative AI features in Discover. The Search version breaks impressions down by page, country, device and date. It does not include clicks, click-through rate, queries or average position, so it measures visibility inside AI answers rather than traffic from them.
Does Search Console show AI Mode clicks?
No. As of September 2026 the generative AI performance report contains impressions only, with no click or CTR data for AI Mode or AI Overviews. The nearest available substitute is segmenting AI assistant referrals in your analytics platform, which captures traffic that actually arrives rather than citations that appear.
When did the generative AI report roll out to all sites?
Google introduced the generative AI performance reports in Search Console in June 2026 and completed the worldwide rollout to all properties on 31 August 2026. Every Search Console property now has access regardless of site size or country.
Are AI Overview impressions counted in the main Search performance report too?
Yes. Impressions from AI features are included in the overall Search performance report totals as well as appearing in the dedicated generative AI report. This is why many sites saw impressions rise while clicks stayed flat: part of the impression growth was AI feature visibility sitting inside the headline number.
My AI impressions are rising but clicks are falling. Is that bad?
It is the expected pattern rather than a fault. Being cited in an AI answer resolves many queries in place, so citations and clicks do not move together. The useful response is to change what those pages are for: optimise them for accurate citation, and put the conversion ask on pages that serve comparison, pricing and implementation queries, which AI answers resolve less completely.
Should I use the control that opts my site out of AI features?
For most commercial sites, no. Opting out does not redirect those users to your organic listing; it removes you as a source while the AI answer is assembled from other sites. Google says the control is not used as a ranking signal outside generative AI features. It is worth considering mainly for publishers whose revenue depends on the page view itself, or where licensing or compliance restricts reuse.
The takeaway
Treat this report as a citation metric with a measurement job attached, not as a new traffic channel. Spend twenty minutes recording four numbers — AI impressions, AI impression share, top-five page concentration, and the clicks and impressions pair beside them — then repeat it monthly. Read the movements through the four patterns rather than as a single “AI traffic” line, pair the report with an AI referral channel in analytics for the half Google will not show you, and improve the pages Google has already told you it finds quotable. The sites that will be able to explain their 2027 search performance are the ones that started writing these numbers down in 2026.
Sources & further reading
- Google Search Central, "Introducing Search Generative AI performance reports in Search Console"
- Search Console Help, "Generative AI performance report (Search)"
- Search Engine Land, "Google Search Console AI performance reports and Search generative AI control rolling out globally"
- Ahrefs, "AI Overviews reduce clicks" (updated study)
- Pew Research Center, behavioural study of clicks on search results with and without an AI Overview