Google explains why AI Overviews don't always appear in search results

Google removes its AI-generated summaries when users don't find them useful, according to Robby Stein, Vice President of Product at Google Search. In an interview with CNN, he explains how a learning system adapts the display of AI Overviews based on user engagement.

A system that learns from user engagement

AI Overviews do not operate according to a default display. Robby Stein explains that the system gradually learns where these summaries add value. When a user asks a short question or searches for a specific website, the AI Overview generally does not appear because it would not be useful in those particular cases.

" The system actually learns where these suggestions are useful and will only show them if users have interacted with them and judged them relevant (...) For many queries, such as short questions or searches for very specific websites, these suggestions will not appear because they are generally not useful. »

The example Stein gives illustrates this mechanism well: when someone searches for an athlete's name, they are generally looking for photos, biographical information, and links to social media. The system has learned that for this type of query, people do not really interact with the AI Overview. Google has many metrics to measure this interaction and adjusts the display accordingly.

Broader background searches

Google is not limited to your exact query. The system often performs additional background searches to broaden your initial request and deliver the most relevant information. This approach explains why some pages appear in AI Overview citations even without exactly matching the keywords you typed.

Content is selected to answer related subquestions or provide additional context. The adaptation is done based on the nature of the query : for image searches, AI Overviews integrate with visual results, while for shopping queries, they connect to product information.

The role of AI mode for complex questions

AI mode represents the next step for complex questions requiring in-depth exchanges. Google designed this mode to allow users to dig into complex topics, such as comparing cars or researching backup power solutions.

During tests of AI mode, Google observed a two- to threefold increase in query length compared with traditional searches. Users also began to ask follow-up questions in a more natural conversational pattern.

Queries in AI mode are becoming much more specific. Instead of simply searching for " things to do in Nashville ", users are, for example, asking " restaurants in Nashville if a friend has an allergy, that we have dogs, and we want to sit outside ".

Deliberately limited personalization

Some personalization already exists in AI mode. Users who regularly click on video results may see those pieces of content ranked higher in their results. However, Stein notes that this personalization remains a minor adjustment to the overall experience.

Google favors consistency across all users while allowing some individual adaptations when they make sense. The goal is to maintain a predictable and consistent experience.

The impact on the presence of AI Overviews

In July 2024, a study showed that Google had reduced the presence of AI Overviews by 52%, dropping from a widespread appearance to only 8% of queries. Stein's explanations offer a clue to understanding this development: the fluctuations reflect user behavior more according to the types of questions rather than actual algorithm changes.

This approach based on real usefulness changes the game for SEO. The query expansion system means that your content can appear in citations even without matching the original wording exactly. This mechanism becomes particularly important for complex queries where AI Overviews are more likely to appear.

The article “Google explains why AI Overviews don't always appear in search results” was published on the site Abondance.