Google AI Overviews Cite Fewer Top Sources

A significant shift is occurring in how Google’s AI Overviews select their sources, with new data revealing a much weaker connection to traditional top search results. Recent analysis indicates that ranking in the top 10 organic results is no longer a strong predictor for earning a citation within these AI-generated summaries. This evolution suggests that content strategies focused solely on keyword rankings may need to adapt to a more complex landscape where topic authority and coverage across related queries are increasingly vital.

An updated study examined millions of keywords and URLs, finding that only 38% of pages cited in AI Overviews also appeared in the top 10 search results for the same query. This marks a dramatic decrease from a similar analysis conducted just months prior, which reported a 76% overlap. The remaining citations are now almost evenly split between pages ranking from positions 11 to 100 and those that do not rank within the top 100 results at all.

Researchers attribute this substantial change to two primary factors. First, improved technical detection methods now capture more citations than before, meaning some of the apparent drop reflects better data collection rather than a complete shift in Google’s behavior. Second, and more critically, is the growing influence of Google’s “query fan-out” process. When an AI Overview is generated, the system can break the user’s original search into multiple related sub-queries. It then identifies and cites sources that appear most frequently across the results for all those sub-queries, not just the main keyword.

This mechanism means a page might be cited for a broad topic even if it doesn’t rank highly for the specific phrase the user typed. The content simply needs to be a prominent source across a cluster of related questions. This analytical approach was bolstered by a global upgrade to the Gemini 3 model in January, which may have enhanced the AI’s ability to perform this kind of contextual sourcing.

Further underscoring the trend, a separate analysis using a different methodology reported an even lower top-10 overlap of approximately 17%. The convergence of these independent studies points to a real and rapid change in how source material is curated for AI answers.

YouTube has emerged as a particularly dominant source, especially for citations that fall outside the top 100 organic rankings. Data indicates that YouTube URLs constitute over 18% of citations from beyond the top 100 and account for 5.6% of all AI Overview citations. It is now reported as the single most-cited domain overall in these features, with its presence growing by more than a third in the past half-year.

This prominence is notable in areas like health queries, where YouTube videos have outranked official medical websites in certain studies. The pattern suggests Google’s AI is actively pulling from video content to answer questions, even when text-based organic results for a query don’t prominently feature video links.

For creators and marketers, the implications are clear. Relying on a high ranking for one target keyword is an increasingly fragile strategy. Success now depends on a page’s ability to serve as a comprehensive resource that answers a wide range of related questions. Content must demonstrate topical authority across a spectrum of user intents, not just match a specific search phrase. The speed of this change is notable, and if the fan-out query process continues to grow in importance, the disconnect between traditional SEO rankings and AI citation could become even more pronounced.

(Source: Search Engine Journal)