Pro-conspiracy search terms can push Google's AI Overviews toward misleading or conspiratorial summaries, according to new research published in the journal Media International Australia. The study focused on two conspiracy theories: chemtrails and 15-minute cities. Researchers collected data from the first page of Google search results over one week in January 2026, comparing AI-generated summaries triggered by conspiratorial search queries against summaries triggered by neutral ones on the same topics.
The findings show that search queries reflecting conspiratorial beliefs returned notably different AI summaries compared with queries that did not include pro-conspiracy language. The effect was strongest when the search query involved lesser-known narratives or used words suggesting the searcher already believed in the conspiracy. For example, searching for "15 minute city segregation" produced a different AI Overview than simply searching "15-minute cities."
Chemtrails is a conspiracy theory that has been around since the 1990s. Believers generally interpret the condensation trails left behind by aircraft as evidence of large-scale weather or climate modification programs. Google was previously called out for giving visibility to conspiratorial beliefs around chemtrails in 2015, and the company responded by demoting problematic search results.
The 15-minute cities conspiracy is newer. The term originally described an urban planning model built around walkable neighborhoods. Conspiracists, however, have reframed it as a scheme by a globalist elite to monitor and control residents through technology.
The researchers developed search queries by drawing on open forums and Google Trends for examples of how people actually search for these topics. They then compared AI Overviews, the sources those overviews linked to, and the conventional ranked list of search results across both conspiracy theories and both types of queries.
The news was not entirely negative. For well-established cases such as chemtrails, the AI Overviews mentioned the conspiracy theory very rarely. Generally, across both topics, AI summaries generated from conspiratorial search terms tended to debunk conspiratorial claims rather than endorse them. However, the research found that the picture was more complicated for lesser-known narratives and for queries where the phrasing itself signaled belief in the conspiracy.
The study's authors called for stronger guardrails to prevent the spread of dangerous conspiracy theories through AI-generated search summaries. AI Overviews are now appearing not only in Google but also in other search engines, and users are increasingly likely to stop reading beyond the AI-generated text at the top of the results page. That makes the accuracy and framing of those summaries more consequential than ever. The researchers note that because users may see only the AI-generated response and nothing else, a single misleading or ambiguous summary carries outsized influence.
