Search Console Still Mixes AI Overviews Clicks With Traditional Search, Google Confirms

There's a recurring expectation among people who handle technical SEO: since AI Overviews have changed click behavior on the SERP so much, Search Console should offer a native way to separate "what came from the AI summary" from "what came from traditional search." The official Google Search Central documentation, updated on December 10, 2025 ("AI features and your website"), settles this question, but not in the way most people expect: that filter doesn't exist. And understanding why it doesn't exist is more useful than imagining that it does.

What the Documentation Says, Literally

The passage is direct: sites that appear in AI features, such as AI Overviews and AI Mode, "are included in the overall search traffic in Search Console," and are reported within the Performance report, specifically under the "Web" search type. There is no separate "AI Overviews" search type apart from the "Web" search type in the way Console currently segments data (Web, Image, Video, News). Clicks and impressions generated when your page appears as a supporting link inside an AI summary add up to the same numbers you already look at for classic search results.

This means that when you open the Performance report and filter by a specific query, the click and impression numbers that show up already include any share coming from AI Overviews or AI Mode, without distinction. There is no additional filter dimension (like the ones that exist for country, device, or search appearance) that isolates "source: AI summary."

Why This Isn't a Bug, It's an Architecture Decision

Google's logic, according to the doc itself, is that AI is "built into Search" and is "integral to how Search functions." In other words: for Google, AI Overviews isn't a channel parallel to search, it's a presentation layer within the same search. The access control a site has over how it's crawled to appear (or not) in these features is the same as always: robots.txt directives for Googlebot. To limit what's shown from your pages, the tools remain nosnippet, data-nosnippet, max-snippet, and noindex (the same tags that already govern traditional search snippets). There is no specific opt-out tag just for AI Overviews.

The practical consequence is that, technically, eligibility to appear as a supporting link in an AI Overview or in AI Mode depends on exactly the same requirements as always: the page needs to be indexed and eligible to appear on Google with a snippet, meeting Search's standard technical requirements. The documentation is explicit in saying that "there are no additional technical requirements" and that there's no need to create machine-readable files, AI-specific text files, or special markup. No new schema.org. Anyone already following good SEO practices (open crawling, internal links, good page experience, accessible text content, structured data consistent with visible text) is already doing what's needed.

What Can Be Inferred Even Without a Dedicated Filter

If Search Console doesn't separate the metric, the realistic approach is to work by approximation, and the documentation itself points to three paths, without detailing the step-by-step: understanding how AI Overviews and AI Mode are counted within aggregated data, analyzing traffic changes over time, and combining Search Console data with Google Analytics.

In practice, this translates into observing indirect patterns:

• CTR drop with stable or rising impressions: if a specific query starts getting more impressions but CTR drops without a change in average position, it's a sign (not proof) that an AI Overview has started appearing for that search and is capturing attention before the click.
• Comparing time windows around known SERP changes: tracking mid- and long-tail queries that historically converted well and cross-referencing them with public announcements of AI Overviews expanding into new verticals.
• Session quality in Analytics: the doc itself states that Google has "seen" ("we've seen") that clicks coming from result pages with AI Overviews tend to be of "higher quality," in the sense that users spend more time on the site. This isn't a Search Console number, it's a Google observation passed along in the documentation, and it reinforces why it's worth cross-referencing with engagement time and conversion in Analytics instead of waiting for this data in Console.

None of these techniques produce an isolated, reliable metric of "clicks from AI Overviews only." They're proxies, and should be treated as such in any internal report: you can raise a hypothesis, not nail down attribution.

What Changes for Those Who Decide Editorial Direction and Schema Based on This

The most direct editorial implication: anyone waiting on Search Console to justify a budget for AI Overviews-specific optimization will keep going without that isolated number for now. This doesn't invalidate GEO work, but it changes the decision criteria: instead of measuring "how much traffic came from the AI summary," the path is to measure what's already measurable (position, impressions, aggregated CTR, post-click engagement) and use sharp shifts in these aggregates, tied to queries where AI Overviews is publicly known to be active, as an indirect signal.

It's also worth remembering the "query fan-out" technique described in the same documentation: both AI Overviews and AI Mode can trigger multiple related searches across different subtopics and data sources to build the answer, and the models identify additional supporting pages during this process. This is a clue as to why a page might gain impressions for a query it had no obvious keyword relationship with: it may have been pulled in by a sub-search generated by the fan-out, not by the original query the user typed. This reinforces the logic that slicing this traffic by a single query, the way Search Console exposes it today, has a structural limit: the source of the click may not match the query shown in the report.

The Separation That Does Exist: Training Control, Not Measurement Control

One point that tends to cause confusion: Google-Extended is mentioned in the documentation, but it serves to limit the use of content in AI training and in other Google "grounding" systems, and it is completely distinct from indexing control for Search. Blocking Google-Extended doesn't remove the page from AI Overviews nor does it affect the Performance report; anyone who wants to control appearance in Search still depends on robots.txt, noindex, and snippet tags.

The honest reading of December's update is this: Google standardized the explanation of how these features enter the numbers, but chose not to open a new filter dimension. Before promising the content team a report on "AI Overviews performance," it's worth checking the primary source and adjusting expectations: the data that exists is aggregated, the rest is well-grounded inference.