Two studies find query length predicts whether Google shows an AI answer

Search Engine Journal published a long piece by Clara Soteras on 9 September 2026 that draws on two recent studies and on Google's own AI Mode usage data. The article makes a specific claim: that "search intent" has split into more than the classic three categories, and that the form of the query, more than any keyword in it, now decides whether Google returns an AI answer at all.

The numbers below are the article's own, and each of the three sources is named.

Google's own AI Mode data

Soteras cites Google's I/O 2026 disclosure that AI Mode has passed one billion monthly active users. Google's numbers describe the queries too. An average AI Mode query is three times longer than a classical query, follow-up questions are rising 40 percent month over month, and more than one in six AI Mode queries now use voice, image, or video rather than typed text. Google is also renaming search intent: the article says Google's new framing describes AI Mode users as trying to "explore, decide, learn, create, or do", replacing the older informational, navigational and transactional split.

The Spanish AI Overviews study

Soteras co-authored the study with MJ Cachón and University of Barcelona researchers Mari Vállez and Carlos Lopezosa, under the CUVICOM project. It measured which Spanish-language queries triggered an AI Overview.

The strongest predictor was query length. A four-word query triggers an AI Overview 48.1 percent of the time, and three-word and four-word queries together account for nearly 70 percent of all AI Overviews. Question modifiers raise the rate much higher: "why" triggers an AI Overview 92.3 percent of the time, "what" 85.7 percent, "who" 68.4 percent.

The study also found that AI Overviews and breaking news barely overlap. AI Overviews appear in 34.6 percent of evergreen searches and in 1.1 percent of immediate-current-event searches, and they co-occur with the "Top Stories" block only 1.4 percent of the time. On 43.6 percent of the analysed pages, neither an AI Overview nor a Top Stories block appears at all.

The eye-tracking study

The second study, run by Diego Criado, MJ Cachón, and the Laika team, tracked what users actually look at and click on inside an AI-driven page. The finding named in the piece is a gap between attention and action: images, product listings, and the "Explore Further" panel inside an AI Overview capture up to 100 percent of user visual attention and produce 0 percent click-through. The parts that actually get clicks are the AI Overview's own text block at 86.4 percent, and traditional organic links, which still get 95.5 percent.

The reversal is that the side panels the SEO industry spent years optimising for are now scanned and skipped. A team building for AI Mode has to plan for its content sitting inside the AI Overview block itself, and behind an organic link that people still click when it is there.

Source

• How AI Is Reshaping Search Intent, What 2 Studies Reveal (Clara Soteras, Search Engine Journal, 9 September 2026)