Context
A controlled comparative study across multiple verticals, languages and query paraphrases reports systematic differences between generative AI search and traditional Google web search in source mix, diversity, freshness, cross-language stability and sensitivity to phrasing.
What changed
The study provides empirical cross-engine evidence that generative-search source selection is not invariant: source composition and visibility outcomes vary across engines, languages and paraphrases, with the evaluated AI-search systems showing a reported tendency toward earned-media sources over brand-owned and social content.
Why it matters
Visibility measurement should preserve engine, language and query-variant dimensions rather than collapsing them into one visibility score. Source type can also be measured as an observed selection outcome, while avoiding unsupported claims that it is a universal ranking factor.
Visibility OS interpretation
Visibility OS interprets this as evidence for engine-specific and language-aware measurement, not proof of permanent production ranking factors or universal causal GEO tactics.
What remains uncertain
How stable are these source-type and phrasing effects across later model versions, geographies and longitudinal measurements? Which observed differences persist under repeated sampling with uncertainty estimates?