“According to our results, ‘AI Overviews’ are not just an improvement to help users but form part of a
broader strategy to maximize revenue,” says Robin Ng from the EPoS Economic Research Center.
“Our research shows that the algorithm tends to withhold AI-generated summaries for certain
searches. This is mostly the case when users make targeted queries, i.e., looking for a specific
product.”
Targeted vs. exploratory searches
To understand the underlying strategy, the researchers have analyzed more than 15,000 Google
search results. They find that “AI Overviews” appear in about half of exploratory searches in which
users are unfamiliar with a topic. However, for targeted searches, Google tends to withhold AI
summaries; they appear in only 10–20% of cases. “Targeted searches are focused, and users have a
specific objective,” says Ng. “For example, they may look for a specific mobile phone model. In these
instances, the data shows that users often do not get the full, AI-powered picture immediately.
Instead, the algorithm displays traditional search results, including prominent advertising slots.”
Algorithm appears to prioritize what is profitable for Google
The researchers suggest that targeted searches are often commercially valuable for the platform and
that this is why “AI Overviews” are initially withheld. However, they observe that “AI Overviews” do
appear when the traditional results page cannot generate enough revenue from advertising. In this
case, Google synthesizes content from various sources, including ad clients, to redirect user attention
to search queries that are valuable for the platform.
Possible extension of market power by means of AI
“Our empirical findings suggest that Google is adapting to recent advances in AI-generated answers,
while using them selectively to protect its advertising revenue,” says Robin Ng. “Such behavior is also
relevant for the design of regulatory frameworks such as the Digital Markets Act. Additional
transparency requirements may be warranted, for example, in the form of obligations requiring
platforms to disclose when and why they deploy AI-generated content.”