Search Console doesn't separate clicks from AI Overviews and AI Mode: here's what you can actually measure about GEO's effect

The Search Console performance report includes traffic coming from AI features in search, but the data is still lumped into the 'Web' type, with no filter that isolates this segment. Understand what can actually be measured about GEO's effect.

Google Search Central's documentation on "AI features and your website" confirms something that changes how GEO's (Generative Engine Optimization) effect is measured in Brazil: sites that appear in AI Overviews and Google's AI Mode enter the Search Console performance report, but they end up mixed together, with no distinction, inside the "Web" search type. The documentation itself is explicit: these clicks "are reported in the Performance report, under the 'Web' search type," with no mention of any filter that would allow separating them as a distinct segment. Anyone who wants to know whether a content restructuring specifically affected presence in AI-generated answers still has to infer this indirectly, not measure it directly.

For AI Overviews and AI Mode, the current documentation doesn't describe any equivalent segmentation filter: the click is still counted within the "Web" aggregate, along with the rest of the organic traffic.

Why this matters and what stays hidden

Without a dedicated segment, a site can gain (or lose) visibility in AI Overviews without that showing up separately anywhere in Search Console. The click from someone who came from an AI-generated answer counts exactly like any other traditional search click, within the "Web" type. That makes it hard to answer the most basic question for anyone investing in GEO: is my page being cited as a source in AI answers, and is that generating traffic of a different quality than traditional search traffic?

Google's own documentation explains how the two surfaces work underneath: both AI Overviews and AI Mode use a technique called "query fan-out," in which the system fires off multiple internal searches on subtopics and related data sources before assembling the final answer. That's why the links shown in an AI answer tend to be more diverse than the traditional ten blue links: the model is trying to cover different angles of the question, not just repeat the standard ranking.

What can be inferred in practice, without a dedicated filter

Since there's currently no search appearance filter that isolates AI Overview or AI Mode within the performance report, the only possible path is an indirect one:

1. Monitor the aggregate performance report (Performance > Search results) of the "Web" type as a whole, observing variations in impressions and CTR over time.
2. Compare two periods, before and after a relevant content change (restructuring an article into questions and answers, adding structured data, strengthening E-E-A-T with clear authorship), using the report's native date comparison tool, even knowing the number includes all "Web" traffic, not just what comes from AI.
3. Cross-reference that aggregate with Google Analytics, checking whether the site's time on page and conversion rate improved during the period, which can be an indirect indication (not proof) that citations in AI answers are bringing higher-quality traffic.
4. Treat any conclusion as a hypothesis, not an isolated measurement: without a technical segment that separates the click coming from AI Overview from the click coming from traditional search, there's no way to confidently attribute a metric variation specifically to GEO's effect.

The point of caution here is statistical and methodological: even in the aggregate, niche sites or those with low search volume may not have enough volume for any minimally reliable conclusion about a specific content change. Before celebrating or panicking over a CTR variation, it's worth remembering that, without segmentation, that variation may have causes entirely unrelated to GEO.

What the numbers tend to show (and why low CTR isn't a verdict)

In practice, it's reasonable to expect more impressions and a lower CTR when content appears in AI Overviews. That happens because the AI answer already delivers part of the information directly on the results page, so a fraction of people who saw your page's citation don't need to click to resolve their question. A lower aggregate CTR for the site doesn't necessarily mean the content got worse; it may mean it's satisfying simple informational searches directly on the SERP. Since there's no way to separate this effect by appearance type within Search Console, comparing the period's CTR to previous periods' CTR, without segmentation, is the only thermometer available today, and it mixes different intent contexts.

Google Search Central's own documentation brings a qualitative data point worth citing: according to the text, when people click from an AI Overview to the site, "these clicks have higher quality," with a higher likelihood that people will spend more time on the site after arriving. That's a clue for cross-referencing Search Console's aggregate data with Google Analytics: if the site's time on page and conversion rate improved during a period when it's known (through other evidence, such as manual SERP monitoring) that the content started appearing more in AI Overviews, that's an indirect argument in favor of continuing to invest in GEO there, even without being able to isolate the exact number in Search Console.

What remains impossible to measure without a dedicated filter

It's worth being direct about the limits, because promising more than the available data delivers is exactly the kind of mistake this coverage should avoid. Without an AI Overview and AI Mode segment in Search Console, today it's not possible to:

• Know how many clicks, impressions, CTR, and average position came specifically from a citation in an AI answer, separate from the rest of the "Web" traffic;
• Compare a page's performance as an AI source with the same page's performance in traditional search;
• Measure in real time the effect of a content change on presence in AI Overviews, since the data is diluted in the aggregate and subject to the same delay as any Search Console report;
• Know which specific passage of the page was used in the generated answer, since not even a future filter would solve this: the documentation doesn't mention that level of granularity.

And, more importantly for anyone thinking about optimizing specifically to appear in AI: Google's own documentation is explicit in saying there's no additional technical requirement to appear in AI Overviews or AI Mode beyond meeting Search Console's basic indexing technical requirements and following already known recommended SEO practices, such as allowing crawling in robots.txt, having organized internal linking, good page performance, accessible text content (not hidden behind heavy JavaScript without a fallback), and structured data consistent with what's visible on the page. There's no text file for AI, nor an exclusive schema.org markup to get into these features. Anyone promising magic GEO optimization with secret markup is selling something the source itself rules out.

What to do with this in practice

For anyone publishing technical content who wants to track GEO's effect rigorously, even without a dedicated filter, the reasonable approach is: monitor the aggregate "Web" performance report monthly, cross-referencing with Google Analytics to observe time on page and conversion, and supplement that with manual, periodic checks of which business-relevant queries are generating AI Overviews with the site as a source (via manual searches, since there's no automated report for this). The most common mistake will be treating a drop in the site's aggregate CTR as a sign of lost relevance, when in fact it may just be a mix effect: more impressions concentrated in AI Overviews, which naturally have lower CTR, pulling the average down even with a stable organic position. Without a segmentation filter, that distinction remains a hypothesis, not a number that can be precisely isolated.