Introduction
Google’s AI Overviews place AI-generated summaries on search results pages, potentially giving users useful information before they visit a source website. A summary of a randomized experiment reports that US desktop users who saw AI Overviews clicked through to external websites about 40% less often than a comparison group for whom the feature was hidden. The result offers experimental evidence in a debate that has often relied on correlations: does presenting answers directly on the search page change users’ willingness to visit the sites behind them?
Key points
- The test used a randomized comparison. Researchers hid AI Overviews for some US desktop users and compared their behavior with that of users who continued to see the feature. The reported difference is attributed in the summary to the experimental setup, rather than inferred only from an association between AI search and traffic.
- The metric is external clicks. The reported decline concerns clicks from Google search results to outside websites. It does not mean that search usage, total website visits, or traffic from every source fell by the same amount.
- The evidence has limits. The available material does not specify the experiment’s duration, sample size, query categories, or types of sites. It also gives no results for other devices or regions, so it cannot establish that every website will lose 40% of its traffic.
Why it matters
For publishers, content sites, and businesses that rely on search referrals, the finding raises a practical concern: appearing on a results page may not translate into the same number of visits as before. If a user finds enough information in an overview, there may be less reason to open the underlying pages. But the source material does not identify which kinds of content were most affected, nor does it measure revenue or conversions. A reduction in clicks should therefore not be treated as an equivalent reduction in earnings.
For Google, AI summaries can make the search page more useful as a destination in its own right, while sharpening questions about how search traffic is distributed between the platform and the sites supplying information. Further assessment will depend on the experiment’s full scope, differences among query types, and user behavior beyond external clicks. The cautious takeaway is that, in this randomized test of US desktop users, showing AI Overviews was associated with—and described in the summary as causing—roughly 40% fewer clicks to external sites. That is a significant signal, but not a universal prediction for the web.
Source: OSChina
Comments
Checking sign-in status...
Loading comments...