Google AI Overviews Can Give Emerging Conspiracy Theories False Balance

Google’s AI Overviews appear to handle conspiracy-related searches unevenly. Research from Queensland University of Technology (QUT), published in Media International Australia, found that the feature generally pushed back on the established chemtrails conspiracy, but could lend undue legitimacy to the newer conspiracy narrative around 15-minute cities.

That difference matters because a 15-minute city is an urban-planning approach intended to improve nearby access to everyday essentials—not a plan to restrict people’s movement.

What the QUT researchers compared

The researchers examined Google AI Overviews, their linked sources and conventional search-result lists for two cases: chemtrails and 15-minute cities. They also compared searches phrased from a conspiratorial perspective with more neutral searches.

Chemtrails is a long-running false claim that aircraft contrails are secretly dispersing harmful substances. Academic research is unambiguous that chemtrails are not real, even though belief in the conspiracy persists.

The 15-minute-city concept, by contrast, concerns whether residents can reach essential services within roughly 15 minutes by walking or cycling. It is a planning framework focused on proximity and accessibility.

Established claims were more often rebutted—but not always well explained

For chemtrails, AI Overviews mentioned the conspiracy relatively rarely and generally debunked conspiratorial searches. However, the study found that these responses could be superficial, often stopping at a statement that the claim was “misinformation” rather than explaining the evidence against it.

That distinction is important. A concise denial may be preferable to repeating a false claim, but it does little to help a reader understand why the claim is unsupported or how to evaluate similar material encountered elsewhere.

The researchers also found that conventional search results were not free of risk: conspiratorial searches could lead users to links including YouTube videos and Facebook posts that supported the narratives.

The 15-minute-cities case exposed a false-balance problem

The more concerning result involved 15-minute cities. In 12.5% of the relevant AI Overviews examined, the summaries presented false conspiracy claims as one side of a seemingly legitimate debate.

This is a false-balance failure rather than simply a factual error. When an evidence-based account is framed alongside an unsupported allegation as though both carry comparable weight, the format can make the unsupported claim appear more credible.

The QUT researchers’ broader point is that a system designed around a fixed set of well-known conspiracies may miss narratives that are newer, evolving or less widely recognized. Waiting until a false claim becomes prominent enough to trigger tailored protections can mean intervening after it has already spread.

Why linked material matters as much as the summary

An AI Overview is not the whole search experience. Its citations and the conventional results around it can shape what a user reads next. QUT’s analysis therefore looked beyond the generated paragraph to the sources and search results returned for conspiratorial queries.

This makes source quality central to misinformation safeguards. A summary may reject a claim while still sending users into an information environment that repeats, reframes or monetizes it. Evaluating the overview without evaluating the pathways it creates would miss part of the exposure risk.

Google faces broader accuracy and accountability questions

The study lands amid continuing scrutiny of AI-generated search summaries. Google has said it has long applied anti-spam policies and protections to generative Search features and continues to upgrade its spam-fighting efforts.

There is also evidence that an AI Overview can reduce the likelihood that a user investigates beyond the summary: Pew Research Center found that Google users are less likely to click links when an AI summary appears in results. That makes clarity, sourcing and appropriate uncertainty especially consequential when a summary handles contested or misleading claims.

Legal scrutiny is growing as well. In June 2026, Reuters reported that Google would appeal a German court ruling holding it liable for false claims appearing in AI Overviews; Google characterized the case as involving specific, narrow errors and said the overwhelming majority of overviews are accurate. 2 Germany’s media regulator later said AI-generated news summaries and chatbot responses are content created by the providers themselves and subject to German media law. 1

The practical lesson: build guardrails for what is emerging

The QUT research does not suggest that every AI Overview about conspiracy theories fails. Its central finding is more specific: protections that work reasonably well for a familiar conspiracy can be less reliable when a narrative is newer.

A stronger approach would:

• provide a meaningful, evidence-based correction rather than a bare dismissal;
• avoid “both-sides” framing where the evidentiary record is one-sided;
• assess the quality of linked sources as well as the generated answer; and
• continuously update safeguards to recognize emerging conspiracy narratives.

For readers, the useful habit is equally straightforward: treat an AI Overview as a starting point, not a final verdict—particularly when it frames a claim as controversial without showing whether credible evidence actually supports it.