Two research pieces published this week — one on zero-click behaviour, one on GEO measurement — form a coherent argument about where GEO strategy currently breaks down and how to fix it.
The zero-click data: 68 per cent of Google searches now end without a click to the open web, up from 49 per cent in 2019. A Spanish media impact analysis combined with an eye-tracking study provides the activation breakdown by query type: "why" queries trigger AI Overviews 92.3 per cent of the time, "what" queries 85.7 per cent, "who" queries 68.4 per cent. Three- and four-word queries account for approximately 70 per cent of AI Overview activations. Evergreen content appears in 34.6 per cent of AI-triggered searches; breaking news effectively suppresses AI Overviews (1.1 per cent activation rate for time-sensitive queries). The eye-tracking study adds a behavioural layer: the AI Overview block converts at 86.4 per cent attention-to-interaction, while peripheral SERP elements — images, product carousels, knowledge panels — capture 100 per cent of visual attention but generate zero clicks.
The measurement research: analysis across 300,000 keywords found that GEO visibility scores — the primary metric most teams report to leadership — can increase while actual clicks decline by 58 per cent when AI Overviews are present. The root problem is definitional: most GEO tools conflate mentions (a brand name appearing anywhere in an AI response) with citations (a URL actively linked as a source). The two are combined into a single "visibility score" that masks the distinction. A brand that is frequently mentioned but rarely linked is accumulating the appearance of GEO success while its AI-search traffic share is declining.
The combined implications are operational. On content investment: query-type data tells you precisely where AI Overviews activate. Evergreen "why" and "what" content at three-to-four-word query depth is where AI surfaces answers; time-sensitive news is largely invisible to AI Overviews and should be evaluated on traditional organic and referral performance rather than GEO metrics. Long-tail conversational queries — the average AI query is now approximately three times longer than a traditional search — are structurally advantaged in AI retrieval relative to short keyword-optimised content.
On measurement: separate citation and mention tracking in whatever GEO tooling you use. If your current tool does not expose this distinction, treat its aggregate visibility score as an unreliable leading indicator until you can verify it against Search Console AI referral data and analytics attribution. Report to leadership on citation rates and AI-referred traffic, not raw visibility scores. The discipline is at a maturity inflection point — the teams that build outcome-connected measurement frameworks now will have the evidence base to defend GEO investment when scrutiny from finance and leadership increases.