Search Console: how to interpret AI Overviews clicks and impressions in the performance report

Google Search Central's documentation details how Search counts traffic from AI Overviews and AI Mode: everything falls under the "Web" type, and rereading CTR requires other clues, not a separate line.

Google Search Central updated the "AI Overviews and your website" page, part of Search Console's official documentation, to clarify a point that has caused confusion for those measuring SEO performance since AI Overviews and AI Mode began occupying significant space on the results page: how this traffic appears in reports.

The documentation's direct answer is less dramatic than many in the GEO community (generative engine optimization) expect: there is no completely separate line, with its own numbers, for "AI Overviews" at the top of the performance report. Google's text is clear: "Just like the rest of the search results page, sites that appear in AI features (...) are included in overall search traffic in Search Console. They're reported in the Performance report, under the Web search type."

In other words, clicks and impressions generated when your page is cited in an AI Overview or in an AI Mode response add up to the total for the "Web" search type, together with classic ten-blue-links results. This changes the correct way to read the aggregated "Web" type numbers in the Search Console performance report, and that's where the real work for anyone doing technical SEO today lies.

Why you can't isolate a single AI number

Google Search Central's documentation confirms that this traffic falls under the "Web" search type in the performance report, together with classic results, and points to a separate external link within the documentation itself for details on how AI Overviews and AI Mode are counted there, content not described in the text consulted. The available material doesn't describe a separate line or filter, with its own number, for "AI Overviews" within the search appearance tab of the performance report, and the existence and exact workings of that kind of segmentation aren't documented in the source. In practice, this means any reading of AI performance in Search today involves working with the "Web" type aggregate and making indirect comparisons over time, not an officially confirmed breakdown.

To recalibrate CTR targets honestly, the way forward is to compare the behavior of the same query cluster over time, before and after AI Overviews started appearing frequently for those terms, looking at the relative variation rather than an isolated number from a specific filter. If a cluster of informational queries (comparisons, definitions, "how to") has started triggering AI Overviews frequently, it's expected that the average CTR for those specific queries will drop even with stable organic position, because part of the answer has already been delivered on the SERP itself.

Why Google says the click that remains is worth more

The documentation includes a sentence worth quoting because it contradicts the most common fear about AI Overviews: "We've found that when people do click from a search results page featuring an AI Overview, those clicks are higher quality, in that users are more likely to spend more time on the destination webpage."

This is an important reading for anyone who only looks at aggregated CTR in Search Console and concludes that "AI is stealing traffic." CTR can fall and revenue or engagement per session can still rise, because someone who clicks after reading an AI-generated summary has already filtered their own intent before reaching the page. The practical recommendation here is to cross-reference the aggregated "Web" type data in Search Console with behavioral data in Google Analytics (time on page, bounce rate, conversion) for the same set of URLs, instead of treating CTR as the only success metric for queries where AI appears.

Query fan-out changes who gets cited

Another technical point in the documentation that directly affects content strategy is the explanation of "query fan-out": both AI Overviews and AI Mode trigger multiple related searches on subtopics to build the answer, and the models identify more supporting pages in the process. In practice, this means the set of sources cited in an AI response can be more diverse than the ten links that would appear in a classic search for the same main question, even bringing in pages that would never appear on the first page for the literal query typed by the user, but that answer one of the internally generated sub-questions well.

This reinforces a behavior that already applies to traditional SEO but becomes more evident in GEO: content structured around clear subtopics, with H2/H3 headings answering specific questions within a larger theme, has a better chance of being pulled in as a supporting source in some branch of the answer, even without leading the ranking for the main term.

No special markup, but the technical fundamentals are still mandatory

A recurring myth in GEO communities is the idea that there would be a dedicated schema.org markup, an "ai.txt" file, or exclusive technical requirements to appear in AI Overviews. Google's documentation explicitly dismisses this: "You don't need to create new machine-readable files, AI text files, or markup to appear in these features." The requirement is the same as always: the page needs to be indexed and qualified to appear with a snippet in classic Search, which involves unblocked crawling in robots.txt (including checking CDN and hosting infrastructure), good page experience, accessible text content, and structured data consistent with the visible text.

For those who want to opt out of these surfaces without leaving the index, the controls are the already familiar ones: nosnippet, data-nosnippet, max-snippet, or noindex. And to limit the use of content in training other Google AI systems, without affecting display in Search, the route is Google-Extended.

How to apply this this week

The practical step is: in the Search Console performance report, track the evolution of CTR and average position for the "Web" type for query clusters where AI Overviews and AI Mode have started appearing frequently, comparing the same cluster before and after that period (since there's no confirmed filter that isolates these appearances on its own), and document the difference by search intent cluster (informational, comparative, transactional). If CTR drops for informational queries but session time rises in Analytics for the same URLs, the CTR target needs to be rewritten by query type, not kept as a single number for the entire site. And if a set of pages never appears either in these clusters or in classic results for related terms, the problem is probably crawling or basic indexing, exactly as it was before AI existed on the SERP, and the URL Inspection tool remains the first place to diagnose that.