On June 3, 2026, Google quietly added something genuinely useful to Search Console: a Generative AI Performance report showing, for the first time, exactly which pages from your domain appear inside AI Overviews and AI Mode — and how often. The data goes back to May 18. It shows only impressions, not clicks. There is no query breakdown. But even in its limited form, it already overturns one of the central assumptions of AIO strategy.
The assumption: optimize for the keywords your tools flag as AIO-triggering, prioritize your high-value category pages, and you control your AI Overview presence. The data says otherwise. By a wide margin.
What the First Numbers Show
Across a B2B domain measured over the first 35 days of available GSC Generative AI data, the impression distribution by content type is stark:
| Content type | Share |
|---|---|
| Glossary / definitional | 49% |
| Blog / how-to / explainer | 26% |
| Product / category pages | 19% |
| All other | 6% |
| Total | 100% |
Glossary and blog content together account for roughly three-quarters of all AI Overview impressions. Product and category pages — the pages that receive the lion's share of SEO investment, conversion optimization, and paid promotion — account for less than a fifth.
The top-performing individual URL is a glossary entry on a core conceptual term with low commercial intent and negligible CPC. It generated more AI Overview impressions in 35 days than the domain's primary product landing pages combined.
The Measurement Gap
This finding also exposes a structural problem with how AIO exposure has been estimated until now. Standard keyword research tools flag individual keywords as AIO-triggering, typically across a domain's top 30–50 ranking terms. Those flags are real — the keywords do trigger AI Overviews — but they measure only a narrow slice of actual exposure.
Keyword tools are optimized to track high-traffic, high-competition terms — precisely because those terms are where rankings matter most for traditional SEO. But AIO fires most intensely on definitional, informational queries: what is X, how does Y work, what does Z mean. These queries sit at the opposite end of the intent spectrum from high-CPC commercial terms. They appear in almost no competitive keyword set. And they account for the majority of actual AI Overview impression volume.
The implication is not that keyword tools are wrong about the queries they track. It is that the queries they track are the wrong sample. The GSC Generative AI Performance report, launched June 3, is currently the only source that measures this correctly — directly, at impression level, from the source.
Why Definitional Content Wins
The mechanism is not accidental. AI Overviews function as CITATION ENGINES for definitional vocabulary. When a user asks "what is [concept]" or "how does [mechanism] work," the AI Overview must source a definition. It retrieves the most retrievable authoritative answer in its training and retrieval corpus — which is almost never a product page and almost always a page whose entire purpose is to explain a concept clearly, completely, and without conversion friction.
Glossary entries, explainer articles, and how-to content are structurally adapted for this retrieval pattern. They contain the concept term, its synonyms, related mechanisms, and a clean semantic context — everything a retrieval system needs to anchor a citation. A category page optimized for conversion sends mixed signals: product benefits, pricing language, CTAs. It is a poor source for a definition even if it ranks well for a keyword.
AIO presence is an architectural outcome, not a keyword optimization outcome. The domains that dominate AI Overviews are those that have invested in definitional depth across their category vocabulary — not just the head terms, but the full semantic neighborhood: sub-concepts, mechanisms, adjacent definitions, common distinctions.
A category page for a core product will rarely appear in AI Overviews. A well-constructed glossary entry for a concept that buyers encounter on the path to that product will appear constantly. The competitive question is no longer which keywords you rank for — it is which concepts in your category you own the definition of.
The Measurement Shift Required
The practical consequence is a change in what you measure and how you act on it. A keyword AIO flag tells you that a specific high-traffic query triggers an AI Overview. That is useful directional information. It tells you almost nothing about actual impression volume, about which pages are generating that volume, or about the long-tail definitional queries driving the majority of AIO traffic.
The GSC Generative AI report answers a different question: which of your pages actually appear, and how often. That data enables content gap analysis — identifying concepts in your category that buyers query, that competitors partially cover, and that you have not yet given a dedicated authoritative page. Each gap is a slot in the AI Overview citation pool that you are not filling.
The rollout of the GSC report is still incomplete — Google launched it initially for a subset of properties and is expanding gradually. There is no API yet; the data requires a manual CSV export per property. Despite those limitations, it is already the most accurate AIO measurement available — and the gap between organizations using it and those relying on keyword tool AIO flags alone is approximately 3.2 times the reported exposure.
The window to act on first-mover data advantage is, as always, shorter than it looks from the outside.
- 01 Google Search Central — Introducing Search Generative AI Performance Reports in Search Console June 3, 2026. Official announcement of the Generative AI Performance report. Data available from May 18, 2026.
- 02 Google Search Console Help — Generative AI Performance Report (Search) Official documentation. Dimensions available: impressions, pages, countries, devices, dates. No query-level breakdown currently.
- 03 rhinegold Compendium — Earned Media Grounding The complementary concept: third-party platform presence that LLMs retrieve and cite. Connects AIO content strategy to broader GEO architecture.
- 04 rhinegold Compendium — Generative Engine Optimization The strategic framework within which AIO content architecture operates.
Impression data based on 35 days of GSC Generative AI Performance report output (May 18 – June 21, 2026) across a B2B domain. Content type classification by URL path pattern.