AI Overview or Overreach? Google’s Strategic Deployment of Generative AI in Search

In May 2024, Google introduced AI Overviews, which synthesize search results into direct answers on the search engine results page, contributing to the growing prevalence of zero-click searches. Notably, Google does not employ AI Overviews for every search query. The key to understanding this selectivity lies in the heterogeneity of search intent, and we argue that AI Overviews are best understood as a strategic instrument for maximizing average revenue per user. We develop a theoretical framework in which a monopolist search platform decides whether to deploy AI Overviews across queries that differ in their search intent along two dimensions: whether the search is exploratory or targeted, and whether it is monetizable. For each intent type, we derive conditions under which the platform benefits from deployment and generate testable hypotheses. To test these predictions, we construct a novel dataset of over 2,000 Google search queries and 15,118 search engine results page observations, where AI Overviews appeared in 31.2% of searches. Consistent with revenue maximization, we find that deployment patterns vary systematically across intent types: for exploratory queries, AI Overviews are deployed by default and withheld only when organic results already suffice or source quality is too low; for targeted queries, deployment is rare and occurs only when the platform lacks confidence in the organic match. Across all intent types, deployment exhibits an inverted-U relationship with source quality. Our findings provide empirical evidence that AI Overview deployment varies strategically with search intent and that AIOs can be characterized as a novel form of platform self-preferencing, with implications for content creators, advertisers, and regulators concerned with platform market power.