Google’s AI search metrics inflate SEO wins, hide clicks

Google’s September 2026 shift did not arrive out of nowhere; it formalized a reporting model previewed earlier in the year. On June 3, 2026, Google announced dedicated Generative AI performance reports for Search and Discover, and by early September the rollout had become broad enough to be summarized by PPC News as: “Google gives every site AI search reports, minus the clicks and queries,” confirming that the new visibility layer centered on AI-answer impressions while withholding the traditional engagement fields marketers had long used to validate search performance.

At the same time, Google made independent verification harder both inside and outside Search Console. PPC Land reported on August 26, 2026 that “the links no longer show the real site address,” instead routing through wrapped google.com/goto URLs, and Derek Perkins at Nozzle said that “If you want to resolve all the links for a five-page ranking, you need between 500 and 1,000 requests,” while Google was also testing a Search Console control letting publishers include or exclude content from AI Overviews, AI Mode, and generative Discover features.

The distortion starts with how Google maps AI answers into legacy SEO fields. John Mueller said Search Console “still treats entire AI-generated overviews as a single block,” so “trying to squeeze AI answers onto an old 1-to-10 ranking scale no longer makes sense”: if the block sits first, every URL inside can inherit position one, whether it is a prominent citation or buried in a dropdown, and average position can then blend that artificial top placement with weaker organic ranks into a flattering number that does not reflect what users actually saw.

Impressions are warped too, because AI Overviews can expand and preload links before a user meaningfully engages. In August 2025, Robby Stein, VP of Product at Google Search, said the main goal for AI Mode was to “display more inline links,” and that it “added link carousels on desktop.” Search Engine Journal later cited data that “External links in AI Overviews rise to 26%,” but more links do not mean more attention: default links can register impressions as soon as the box loads, expanded answers push classic organic results lower, and some interactions keep users inside Google’s AI flow, leaving reports rich in visibility signals but poor in evidence of real engagement.

Marketers now have to treat AI search measurement as an attribution problem, not a ranking report, because Search Console and analytics each miss different parts of the journey. As Combining Data Sources to Overcome AI Attribution Challenges puts it, “no single tool gives you all 10 metrics,” especially when an AI recommendation can send a click as referral traffic, trigger a branded Google search logged as Organic, or produce a typed-in visit that lands in Direct, giving AI influence zero credit unless teams reconcile those paths across systems.

The practical response is a stitched workflow: use Search Console for AI visibility context, GA4 with custom channel groupings for referrers such as chatgpt.com and perplexity.ai, lead-form fields asking “how did you hear about us?”, server logs to parse AI crawler user agents by page, and prompt and brand-monitoring tools to track mentions, citations, and sentiment. That matters even more because Leadership in SEO warned that “the Pages export doesn’t necessarily represent every page or impression because of the thousand-row limit,” adding that “If you go over the 1,000 row limit, you are seeing a sample,” so business impact has to be inferred from converging signals against a consistent KPI like leads or revenue.