New research finds that Google's AI Overviews can reinforce conspiracy theories rather than debunk them. The study, reported by The Conversation, directly challenges Google's framing of AI Overviews as a feature that helps users understand complex topics with synthesized, sourced answers.

The finding lands at a difficult moment for Google's search product. AI Overviews rolled out across the United States in May 2024 and expanded to more than 100 countries by late that year. Since launch, the feature has drawn repeated criticism for surfacing inaccurate, satirical or harmful content — including widely reported incidents of advice to eat rocks or add glue to pizza, both traced to low-quality sources the system summarized.

What did the research find?

The Conversation reports that the new research tested how AI Overviews handle queries related to conspiracy theories. The results showed the feature does not consistently counter false narratives. In some cases, the generated summaries reinforced conspiratorial claims instead of refuting them.

This is a research finding, not a statement from Google. The study examined the behavior of Google's own AI-generated output — a confirmed product system — but the conclusions come from independent academic analysis rather than from the company. Google has not, in the source material, responded to this specific research.

The mechanism matters as much as the outcome. AI Overviews are generated by Google's Gemini-based systems, which synthesize answers from pages the search engine itself ranks. When those systems pull from unreliable or conspiratorial pages, the summary can present fringe claims with the visual authority of a Google answer box. Users see a synthesized response above the traditional ten blue links, which reduces the likelihood they will check the underlying sources.

Why does this matter for search?

The research speaks to a structural question in modern search: who bears responsibility for AI-summarized answers. A traditional result list presents competing pages side by side. An AI Overview presents one answer, assembled by an automated system, at the top of the page.

That difference carries real weight for publishers and for public information. Sites that produce careful, evidence-based reporting on contested topics — health, elections, science — can find their work compressed into a summary that sits next to, or blends in, lower-quality material. For verticals where misinformation spreads easily, such as medical advice and political claims, the stakes are highest.

Google has repeatedly said it continues to refine AI Overviews and has pointed to its quality systems as a safeguard. The company's public position treats the feature as an evolving product with ongoing adjustments. The new research suggests those safeguards do not yet consistently filter conspiratorial content out of generated summaries.

What should search professionals watch?

For SEO practitioners and publishers, the study adds a data point to a growing body of evidence on AI Overviews quality. It also raises a practical question about source selection. Understanding which sites Google's systems choose to synthesize — and which they exclude — is becoming central to visibility in an AI-summary search results page.

The research contributes to wider scrutiny of AI-generated answers in search. Regulators, academics and publishers have all pressed Google on how AI Overviews attribute, select and summarize content. Studies like this one supply the evidence base for those debates.

The next signal to monitor is Google's response. If the company adjusts how AI Overviews handle conspiratorial queries — through policy changes, ranking changes or new source-quality filters — that will indicate whether independent research of this kind is shaping the product. Until then, the confirmed fact stands: independent analysis has found that Google's flagship AI search feature can reinforce the very claims it should refute.