June 3: GEO Measurement Got an Official Ruler
In early June 2026 (reported June 3), Google added a generative-AI performance report to Search Console. It aggregates URL impressions from AI Overviews, AI Mode, and Discover's generative features across page, country, device, and date dimensions. AI visibility that teams had only been able to guess at through third-party estimates now shows up as an official ruler inside the console. The rollout is staged to a subset of sites, so the report may not be visible in every account yet, and an opt-out toggle that blocks AI impressions ships alongside it.
A Report With No Clicks — What Changes
There is no click data in this report. You can see how many times your page was surfaced inside an AI Overview, but the console never tells you whether that impression turned into a session or a conversion. Unlike the classic search performance report, which pairs impressions, clicks, CTR, and average position, the generative report gives only one official axis: impressions. If your KPIs were built on clicks, the moment you force this data into a synthetic CTR your numbers distort.
In a Zero-Click World, Impressions Are the Baseline
Treating a zero-click impression as "no result" undervalues the whole of your GEO work. An AI Overview imprints your brand and your claims even when the user never visits the page, so you need a structure that makes the impression itself the baseline and tracks downstream indirect conversion separately. Per-page AI impression share, AI-impression-to-branded-search growth, and the lag between impression and conversion are the three axes to build the metric set around.
From Design to Operations: An AI-Impression KPI Rebuild Checklist
(a) Fix the target numbers first. This is not about scrapping click KPIs — it is about layering impression-based metrics on top of them. Track the AI impression share of your top 20 landing pages (each page's slice of total AI impressions) monthly, and declare a quarterly target on two axes, like "combined AI impressions of the 10 core pages up +15% month over month" in absolute volume and in share. Keep the click metrics, but place AI impressions next to branded search volume so you can check by correlation whether branded search rises as impressions grow.
Because official data only starts accumulating in early June, bake the short baseline into your targets. Until at least four weeks have piled up, watch absolute impression volume and page-rank movement rather than trends, and phase in growth-rate targets only after week eight.
(b) Three failure patterns break this most often. First, multiplying AI impressions by a made-up CTR to force them into a click KPI — the console gives no clicks, so a target built on estimated clicks leaves nothing but ungrounded numbers on the dashboard. Second, splicing pre-June third-party visibility estimates onto post-June official impressions in one chart — joining two series measured differently distorts the entire trend. Third, flipping the opt-out toggle with no performance review and cutting off your own impressions.
(b') The recovery branch starts with separating the data series. Annotate the date the official report began so estimates before it and measured values after it never overlap on the same axis, and do not enable opt-out on a page before reviewing at least four weeks of its impression trend and conversion contribution. For accounts where the staged rollout means the report is not visible yet, defer opt-out and KPI changes and switch properties over in sequence as their data arrives.
(c) Run a weekly GEO report on a fixed template. Fill the same table with five items every week: (1) the top 20 pages by AI impressions and their week-over-week change, (2) pages whose impressions plunged (down 30% or more week over week) with a cause hypothesis, (3) cited-paragraph tracking — which h2 or summary paragraph gets cited for representative queries in AI Overviews and AI Mode, (4) shifts in the impression distribution by device and country, and (5) the list of opted-out pages. Since the report exposes page, country, device, and date directly, this table fills straight from a console export.
Build the reporting pipeline to auto-load the console API or a manual export once a week. The minimum fields are page URL, AI impressions, country, device, date, and a data-source flag (official vs. estimate). Keeping the source flag lets you later filter to the official series to redraw a trend, and it blocks the estimate-mixing accident at the pipeline layer.
(d) The improvement loop turns at the paragraph level. Pull the shared structure of pages whose impressions rose (question-style h2, a 40–60 character summary sentence, tables and lists) and fold it into the spec for your next content refresh; for pages whose impressions plunged, diff against the content change history to see whether the paragraph that used to be cited disappeared. Judge whether the loop is really turning by whether core-page AI impression share and the branded-search contribution of rising pages climb together.
Takeaways at a Glance
With official impression data in hand, the job is not to discard click KPIs but to add an AI-impression axis on top of them. Stand up per-page impression share, branded-search growth, and impression-to-conversion lag as metrics; separate the official series that began in June from earlier estimates; and run a weekly report covering top and plunging pages plus cited paragraphs as a checklist — and even a hard-to-reverse decision like opt-out becomes something you can settle with data.
References
Introducing Search Generative AI performance reports in Search Console — Google Search Central Blog
Google finally gives Search Console its own generative AI visibility reports — PPC Land