Google completed the global rollout of generative AI performance reports in Search Console on Sunday, making AI-specific impression data available to every verified property worldwide. The reports — which had been in phased rollout since June 2026 — track impressions generated by AI Overviews, AI Mode, and generative features in Discover, broken down by page, country, date, and device. Simultaneously, Google activated an opt-out control in Settings that allows site owners to exclude their content from AI Overviews, AI Mode, and Discover's generative features without penalty to standard organic rankings — a mechanism that the UK Competition and Markets Authority had separately mandated by March 2027, suggesting the timing is not entirely voluntary.
The rollout represents the first point at which brands can systematically measure AI citation frequency at the property level using a first-party tool rather than third-party sampling methodology. That is a genuine infrastructure development for the GEO discipline. But the critical limitation — absent from the official announcement, present in every practitioner review of the reports — is that click data is not included. Practitioners can see that an AI Overview generated ten thousand impressions for a given page. They cannot see how many of those impressions produced a visit, a click, or any downstream action.
The measurement gap this creates is not accidental. It is the central tension in GEO measurement as a practice: the mechanism by which AI features generate commercial value for publishers is incompletely understood, because the data needed to understand it is not available in standard analytics stacks. The Google Search Console reports close the impression visibility gap and open a new one. Impressions without clicks is, for performance marketing purposes, awareness data — useful for brand strategy, limited for attribution, useless for CPA models. The brands best equipped to use the new reports are those that have already built the adjacent measurement infrastructure: the Brainlabs 335 per cent conversion uplift finding requires Google Analytics event tracking to surface; the RankCaster AI citation lifetime data requires cross-provider monitoring to generate. Search Console's AI reports are a necessary but not sufficient measurement layer.
The opt-out mechanism is worth separate consideration. It gives publishers the first empirical tool to measure the actual traffic impact of AI Overview inclusion versus exclusion at the page level — a question the industry has debated theoretically since AI Overviews launched. The data generated by publishers who choose to test exclusion will be the most commercially useful AI search research of 2026. Expect it to start emerging in Q4.