Search Console AI Reports Fall Short, Google's Mueller Concedes

A rare admission about Google's own measurement tool

In mid-September, Google's Search Advocate John Mueller agreed with a Reddit user's critique of how Search Console reports visibility in AI Overviews and AI Mode. He said that tracking position for these features in a useful way is hard, and that Google currently has no better method.2122 The exchange took place roughly two weeks after the company finished rolling its new Search Generative AI performance reports out to every site worldwide on August 31, 2026.1 The timing makes the admission more pointed: the tool site owners had requested for two years became widely available, and almost immediately the company that built it said it could not answer one of their most basic questions.

The coverage agrees on what happened. Where it differs is in how much weight to put on it. Search Engine Journal described the exchange as Google conceding that the reporting is inadequate.21 RebelMouse went further and said Google had admitted its AI search data is "broken."26 Mueller's own wording was milder. He described the problem as one of usefulness rather than accuracy and asked SEOs to suggest what position tracking they would find helpful.22 The fairest reading is in the middle. The numbers in Search Console are internally consistent, but they describe a search experience that no longer exists in the form the metrics assume.

What the report counts, and what it misses

Google announced the generative AI reports on June 3, 2026, as a separate view of how often a site's URLs appear in AI Overviews, AI Mode and generative features in Discover. Users can break the data down by page, country, device and date.1 Several analyses point out that the view only separates out data Search Console already counted. The AI impressions were always part of the overall performance totals, so the two should not be added together.521 The first version also leaves out clicks, click-through rate and query-level data. It does not separate AI Overviews from AI Mode either, so a site cannot tell which surface produced its impressions.37

The Reddit post that started the discussion identified several ways the counting can mislead.2225

• Impressions without eyeballs. Search Console counts an impression when an AI Overview containing a site's link loads on the results page, whether or not the user ever scrolls to it.22
• The "Show More" exception runs the other way. Links hidden behind the expansion control are not counted until someone clicks to reveal them, so that part of a site's exposure is undercounted.25
• Position belongs to the box, not the link. Every link inside an AI Overview gets the overview's own position on the page. The average position a site sees therefore reflects where the AI block sat, not where its link appeared among the cited sources.21

Mueller confirmed this approach. He said Google tracks AI features "as a block," as it does for many search features, and that position is not separated out in the generative AI report.22 He also said that interactive results pages make the old "position 1–10" model hard to map onto what users actually see.22

The result is a measurement system that can overstate visibility in one place and understate it in another. Its main ranking metric describes a container rather than the content inside it. RebelMouse's analysis gives the practical consequence: a query with high impressions and low clicks could mean a site was outranked normally, was cited in an AI Overview and ignored, or was paraphrased without any link. The dashboard reports all three situations the same way.26

Why this lands harder in the AI Overviews era

This would matter less if AI Overviews were a minor feature. They are not, and outside research on their effect on traffic keeps showing larger impacts. Ahrefs found in 2025 that AI Overviews were linked to a 34.5% lower click-through rate for the top-ranking page.20 When it repeated the analysis with December 2025 data, the figure had risen to about 58%.11 A randomized field experiment by researchers at the Indian School of Business and Carnegie Mellon found that showing an AI Overview cut outbound organic clicks by 39.8% and raised zero-click searches by 34.5%. Clicks on ads did not change.13 Pew's tracking of real browsing found users clicked a traditional result 8% of the time when an AI summary appeared and 15% when none did.12

These studies use different methods and measure different things, so they should not be merged into a single "traffic loss" figure.19 Seer Interactive's long-running data also showed CTR on AI Overview queries partly recovering in early 2026, though it stayed well below queries without an overview.15 Still, every study points in the same direction. AI Overviews substitute for visits to publisher sites, and the effect is concentrated on informational queries where the overview sits at the top of the page.13

That context explains why the weakness of Search Console matters so much. Publishers are working out whether AI features are worth staying in. They now have an opt-out toggle that Google says removes them from AI Overviews, AI Mode and Discover AI without affecting regular rankings.56 Their best first-party evidence for that decision is a report with no click data, no separation between surfaces and no link-level position. A site owner cannot weigh the value of AI visibility against the cost of lost clicks when the clicks are not reported and the visibility figure is unreliable.

It is also worth noting that impressions are the metric Google chose to show. Industry figures cited by Omnibound show total search impressions rising sharply since AI Overviews launched while clicks fell.15 A report built around impressions alone tends to emphasize the growing figure and leave out the shrinking one. This is an editorial observation about how the tool is designed. Nothing in the reporting shows that Google intended it that way.

A reversal, then a ceiling

The report itself marked a change of position. Rich Sanger's analysis recalls that Google previously treated AI Overviews as just another search feature that did not need separate reporting, and that Danny Sullivan said there were no plans to create it.8 Several commentators linked the June launch, which came with the opt-out toggle and started with UK sites, to pressure from the UK Competition and Markets Authority.5 Publishers have also escalated: Penske Media filed an antitrust suit and the European Publishers Council complained to the European Commission.14

Mueller's comments suggest Google has reached the limit of what its current reporting model can do. The company accepted that AI visibility deserves its own category. It then filled that category with the same impression and position concepts it uses for blue links. Search Engine Journal argued that the old model has not reflected reality for years, yet Google keeps fitting AI results into it.21 A later piece from the same outlet advised SEOs to stop focusing on position and measure outcomes instead, since Search Console flattens each AI Overview into a single block.29 RebelMouse pointed out another tension. Google is asking publishers to fact-check AI-assisted content more carefully, while its own tools give them little reliable information about what happens to that content in AI search.30

What practitioners are doing instead

The practical advice across the coverage is consistent. SEO Vendor recommends treating the generative AI figures as directional signals rather than proof at the page level, checking them against sessions, leads and conversions, and avoiding budget decisions based on AI exposure alone.23 RebelMouse recommends server log analysis to see which AI crawlers visit which pages, and warns against reporting AI citation counts as a headline KPI.26 Reliqus notes that Search Console does not cover ChatGPT, Perplexity or other non-Google AI engines, so a complete picture requires third-party monitoring tools combined with analytics data.7

That last point puts Google's problem in perspective. Search Console only sees one company's AI surfaces, and even for those it now admits the picture is incomplete. Our view is that the September exchange matters less as a confession about one dashboard than as confirmation that AI search has outgrown the metrics the SEO industry built itself around. Mueller asked for ideas, and Google has said more metrics may come.322 Until it adds clicks, separates the surfaces and reports where a link actually sits, site owners should treat the generative AI report as one rough indicator among several, not as an accurate record of how they perform in AI search.