Search Console’s AI report: did AI take your clicks?
The report alone can’t tell you. It counts impressions (how often AI Overviews and AI Mode showed a link to your site), not clicks. Your AI clicks are already counted in the Performance report, under the Web search type, and no filter separates them. To size an AI loss, compare pages with high and low AI exposure.
What does Search Console’s AI report show?
Google launched the Generative AI performance report on June 3, 2026, for a subset of sites, and says it reached all websites worldwide on August 31, 2026 (Google Search Central). You’ll find it in Search Console under Performance.
It counts one thing: impressions in AI Overviews (the AI summary above the results) and AI Mode (Google’s chat-style search), added together. Read it as exposure: how often an AI answer showed your pages. It can’t tell you whether anyone clicked.
| Data | In the AI report | What to know |
|---|---|---|
| Impressions | Yes | AI Overviews and AI Mode combined into one number. Search Labs experiments are left out. |
| Pages, countries, dates, devices | Yes | The four tabs above the table. Dates are in Pacific Time. |
| Search type | Yes | Web: text-based, or Web: multimodal (searches made with an image, such as Google Lens or Circle to Search). |
| Clicks and CTR | No | AI clicks are counted inside the Web search type of the Performance report, with no way to separate them. |
| Queries | No | Only the Performance report has queries, with AI answers and blue links mixed together. |
| Position | No | In the Performance report, a whole AI Overview counts as one position. |
| AI Overviews and AI Mode apart | No | No Search Console report splits them. |
| Discover | Separate report | The Discover version has pages, countries and dates, but no devices. Don’t add it to the Search numbers. |
| API | No | Export button only. The Search Analytics API and the bulk data export list no AI search type (checked October 23, 2026). |
Where are your AI clicks?
In the Performance report, under the Web search type, mixed in with your ordinary clicks.
Google says sites that appear in AI Overviews and AI Mode are counted in overall search traffic and reported “within the ‘Web’ search type” (Google Search Central, updated December 10, 2025). A click on a link inside an AI Overview or an AI Mode answer counts as an ordinary click (Search Console Help). The AI report itself is a filtered view of that same Web data. So:
- Don’t add AI impressions to Web impressions. They’re already included.
- Don’t divide clicks by AI impressions. The clicks include every blue link, so the result isn’t an AI click-through rate.
- Analytics can’t separate them either. A click from an AI Overview shows up as Google organic traffic or with no referrer at all (Ahrefs, January 26, 2026).
How does Google count an AI impression?
An impression is one time a link to your site was shown in an AI feature. Google applies its standard impression rules to both AI Overviews and AI Mode (Search Console Help):
- Served, not seen. A link counts when it’s on the results page, whether or not the person scrolled down to it.
- Links behind a click count later. If the person has to click to reveal a link, such as “Show more”, it counts only once they do.
- One per site per answer on the chart. If one AI answer links to two of your pages, the chart counts one impression. The Pages table counts one for each page, so its rows can add up to more than the chart total (Search Console Help).
- Final, canonical pages. The Pages tab groups impressions by the page the AI answer linked to, after any redirects, and most data goes to its canonical URL (the version of a page Google treats as the main one).
So the number overstates exposure in one way (served links nobody saw) and understates it in another (links behind “Show more”). Use it to compare your own pages with each other. It isn’t a count of people who saw you.
Position works differently, and it matters for the next section. Google’s rule is that “an AI Overview occupies a single position”, and every link inside it shares that position. The position Search Console reports for a query is your topmost link (Search Console Help).
Three traps before you compare anything
The impressions logging error
Google says a logging error stopped Search Console from reporting impressions accurately from May 13, 2025 until April 27, 2026. Impressions, CTR (click-through rate: clicks divided by impressions) and average position were affected. Clicks were not (Google’s data anomalies page, entry dated April 3, 2026).
Google says impressions may look lower now that the error is fixed, which suggests the error inflated them. Inflated impressions push CTR down, so a CTR comparison across the fix can shrink or hide an AI-driven fall. If either of your date ranges touches that period, compare clicks only.
A second, smaller error lowered impressions in the AI report itself from August 13 to 17, 2026. Google restored the data on August 21, so export that week again if you saved it before then.
An AI Overview can hide a ranking fall
Because Search Console reports your topmost link, a page cited in an AI Overview at the top of the results can show a position near 1 even if its blue link slid from 3 to 8. Average position can look steady while the ranking you care about fell. That is our reading of Google’s definitions; Google doesn’t spell it out. Check organic rank with a rank tracker, or search the query yourself while logged out.
AI Mode follow-ups are new queries
When someone asks a follow-up question in AI Mode, Google treats it as a new query, with its own impressions, position and clicks (Search Console Help). Site owners have spotted them in the Queries table as fragments like “yes” and “yes go on” (Search Engine Roundtable, August 6, 2026). Their CTR and position mean little and they distort query-level comparisons. Filter out one- and two-word replies before you compare queries.
How much of your drop did AI cause?
Check your redesign and Google first. Broken redirects, a leftover noindex tag or a Google update can look like AI, and the first two are fixable. Start with your redesign and Google. If your positions fell, stop here: AI answers take clicks while rankings hold.
Pick two equal date ranges, and compare clicks. After: the latest 28 full days, ending a few days ago so you skip the newest data, which Google marks as preliminary. Before: the same number of days, starting on the same weekday, before the drop. For a seasonal business, use the same weeks a year earlier. If either range touches May 13, 2025 – April 27, 2026, don’t compare impressions or CTR between them; compare clicks.
Export by page. From the Performance report (Web), export clicks, impressions and position by page for both ranges. The interface shows 1,000 rows; the Search Console API returns far more. From the AI report, export the Pages tab for the after range, by hand. If your drop began before mid-May 2026, you can measure today’s AI exposure but not the exposure at the time of the drop.
Work out each page’s AI share. Divide its AI impressions by its Web impressions for the same range. Use the result to rank pages against each other. It isn’t an exact percentage, because AI impressions are counted differently.
Keep pages you can compare. Same page type (guides with guides, product pages with product pages); at least 100 clicks in the before range, so ordinary swings don’t dominate; position within one place of where it was; Web impressions within about 15%, so demand held. Skip the impressions test if the before range falls inside the logging error.
Compare high with low, and report a range. Split the pages into high AI share (30% or more) and low (under 10%). For each group, total the clicks: after divided by before, minus 1, is the group’s change. The difference between the two changes, times the high group’s before clicks, is the estimated AI loss. The low group stands in for everything that hit both groups, such as season, demand and tracking. Statisticians call this a difference-in-differences. Re-run it with other cut-offs (40% and 5%, then 25% and 15%) and report the lowest and highest results.
Check the queries. The AI report has no queries, so use the Performance report. Take the top queries of your biggest losing pages, search them logged out, and note which show an AI Overview. Compare their click-through rate at a steady position with similar queries that show none. Leave out brand searches and AI Mode fragments.
Put a value on it. Multiply the lost clicks by your conversion rate and the value of a conversion, from your analytics. Give a range, as with the clicks.
| Group | Pages | Clicks before | Clicks after | Change |
|---|---|---|---|---|
| High AI share (30% or more) | 14 | 10,000 | 6,800 | −32% |
| Low AI share (under 10%) | 22 | 8,000 | 7,200 | −10% |
In this example, the high-AI pages fell 22 points more than the low-AI pages. 22% of their 10,000 before clicks is about 2,200 clicks per 28 days: the estimated AI loss. The other 1,000 clicks they lost are the 10% that the low-AI pages lost too, so something other than AI took those. Re-run with other cut-offs and the estimate might land anywhere from 1,700 to 2,600. That range is the answer to give, not 2,200.
How sure can you be?
Label every finding with one of three words before you show it to anyone.
| Label | What belongs there |
|---|---|
| Verified | Clicks fell. Position and impressions held. The pages that lost the most have a high AI share, and low-AI pages of the same type held up better. Search Console shows all of this directly. |
| Inferred | That the gap between the two groups is AI, and how big it is. Pages with heavy AI exposure tend to be question pages, which can shrink for other reasons. Comparing pages of the same type reduces that risk. It doesn’t remove it. |
| Unknowable | Clicks lost to ChatGPT, Perplexity, Gemini and other assistants. The split between AI Overviews and AI Mode. Whether anyone looked at a served AI link. Your AI exposure before the drop, if the drop came before the report’s data starts (mid-May 2026 at the earliest). |
What does the research say?
| Source | Published | Design | Finding |
|---|---|---|---|
| Pew Research Center | July 22, 2025 | Browsing data from 900 US adults, March 2025. Observational. | People clicked a traditional search result on 8% of visits with an AI summary on the page, and 15% without one. They clicked a link inside the summary on 1%. |
| Ahrefs | February 4, 2026; updated October 15, 2026 | 300,000 keywords, December 2023 against December 2025, aggregated Search Console data. A correlation. | The top result’s click-through rate on keywords with an AI Overview was 58% below a forecast built from keywords without one. A modelled gap, not an experiment. |
| Agarwal and Sen (Indian School of Business, Carnegie Mellon) | April 3, 2026 (working paper, SSRN) | Randomised: a browser extension hid AI Overviews for some of about 1,065 US desktop users, January–February 2026. | On searches with an AI Overview, hiding it raised organic clicks from 0.37 to 0.62 per search: about 40% fewer clicks with the overview. Not peer reviewed. |
| Wang, Gleason, Bart, Wilson and Metaxa | August 18, 2026 (preprint, arXiv) | Randomised and preregistered: 1,100 US Chrome users, recruited March 2026, 7 days in the assigned condition. | Forcing every search into AI Mode cut click-through to websites by 18.8 percentage points. Forced use for a week, not normal behaviour. Not peer reviewed. |
| Google (Liz Reid, Head of Search) | August 6, 2025 | Company statement. | Says total organic clicks from Google to websites have been “relatively stable” year over year, with slightly more “quality clicks”. Published no figures. |
The experiments say an AI answer cuts clicks to websites on the searches where it appears. Google’s claim is about total clicks across all searches, and AI answers show on only some of them: 18% of searches in Pew’s data, about 41% in Agarwal and Sen’s. Both can hold at once. That’s why a sitewide total tells you less than a page-level comparison.
Should you exclude your site from AI features?
Search Console has a control for it: Settings, then Search generative AI. You can include your site (the default), exclude it, or inherit the setting from a parent property (Search Console Help).
Excluding stops your links and content from appearing in AI Overviews, AI Mode and AI features in Discover, and stops your content being used to build their answers. You get no impressions or traffic from them. Google says it takes effect 1–2 days after the change goes live, longer for some content.
What it doesn’t do, by Google’s account:
- It doesn’t remove the AI answer. Content from other sites, which may look similar to yours, still appears.
- It isn’t used for ranking. The control isn’t a ranking or inclusion signal for the rest of Search.
- It doesn’t stop AI training. That’s a separate robots.txt token, Google-Extended.
- It doesn’t work page by page. It covers the whole property. For single pages, the older preview rules (
nosnippet,data-nosnippet,max-snippet) limit what Google shows from a page across Search, AI features included, and your ordinary snippets with it (Google Search Central).
When it’s worth it: rarely, if your problem is lost clicks. Excluding gives up the clicks AI links still send, and the AI answer still shows, built from other sites. It makes more sense when being summarised costs you more than those links bring in, for example when the answer itself is what you sell. Look at the AI report first: Google suggests it for judging how a change would affect your traffic.
No report, no API: what to do
If you can’t see the report
Google gives two reasons: your site hasn’t had enough impressions in AI features, or it’s excluded through the control (Search Console Help). A property that covers part of a site, such as one folder, inherits the control from its parent by default, so check the domain property too. Google’s help page for the Discover version still says that report is rolling out over time (read ).
Export it every month
There’s no API for the AI report: the Search Analytics API and the bulk data export list no AI search type (checked October 23, 2026). The Export button downloads the chart and the table, and values shown as ~ or - come out as zeros.
Google hasn’t said how far back the report goes. On the site he previewed it on, SEO consultant Brodie Clark saw data starting on May 18, 2026 (his preview, June 5, 2026). Export the Pages tab once a month, or you’ll have no baseline when you need one.
What’s coming
The UK’s Competition and Markets Authority requires Google to give publishers metrics for its AI search features, including impressions, clicks and click-through rate, and has set December 3, 2026 as the deadline (CMA summary, August 4, 2026). The requirement covers Google’s AI features in the UK. Google hasn’t said whether the clicks will reach Search Console, or whether sites outside the UK will get them. We’ll revise this guide when Search Console adds AI click data.
What this means for your drop
Your positions fell.
Then it isn’t only AI: AI answers take clicks while rankings hold. A ranking fall points to a Google update or a site problem. Check your drop date against Google’s confirmed updates with the free Google update checker.
GoogleYou relaunched or moved before the drop.
Broken redirects and leftover noindex tags cut clicks in ways that can look like AI. Test your old URLs with the free redirect checker and key pages with the noindex checker.
Your redesignRankings held, and your high-AI pages lost the most.
That’s the AI pattern. Those clicks are structural: fixing your site won’t bring them back, so we sell no fix for them.
AI
The Autopsy answers this question with your Search Console data, after ruling out your redesign and Google. Its AI exposure map shows which of your lost queries now show an AI answer, and who gets cited instead. Not sure where to start? The free Cause Check names the leading suspects from outside data, by email within one business day.
Frequently asked questions
Where is the AI report in Search Console?
Under Performance, as the Generative AI report for Search, with a separate one for Discover. Google’s help page for the report links straight to it. If it isn’t there, your site hasn’t had enough impressions in AI features yet, or it’s excluded through the Search generative AI control.
Does Search Console show clicks from AI Overviews?
It counts them, but you can’t see them on their own. Clicks from AI Overviews and AI Mode are in the Performance report under the Web search type, mixed with every other click. The AI report shows impressions only.
How do I track AI Mode traffic?
Not directly. AI Mode impressions are in the AI report, combined with AI Overviews. Its clicks are in the Performance report, mixed with everything else, and its follow-up questions show up as separate queries. Track AI exposure by page in the AI report and clicks for the same pages in the Performance report.
Can Google Analytics show clicks from AI Overviews?
Not separately. A click from an AI Overview arrives as Google organic traffic or with no referrer, so nothing marks it as coming from an AI answer.
Why don’t the chart total and the table total match?
The chart counts one impression per site per AI answer. The Pages table counts one per page. If an answer links to two of your pages, the table counts two and the chart counts one.
Will excluding my site from AI features hurt my rankings?
Google says no: the control isn’t used as a ranking or inclusion signal for other parts of Search. You do give up the impressions and clicks that AI features send you.
Do AI Overviews reduce website traffic?
On the searches where they appear, the evidence says yes. Two randomised experiments found fewer clicks to websites when an AI answer was shown: about 40% fewer with an AI Overview, and a click-through rate 18.8 percentage points lower in AI Mode. Whether your own drop came from them is what the page-level method on this page estimates.