On January 2, 2026, The Guardian published an investigation finding that Google’s AI Overviews — the generative summaries pinned to the top of search results — deliver false or misleading health information. UK charities including Pancreatic Cancer UK, the British Liver Trust, The Eve Appeal, and Mind warned that people are being put at risk of actual harm.
An AI Overviews failure is not news by itself. What sets this one apart is the specificity of the evidence: every error comes with a disease, a query, and a named expert vouching for why it matters. For any team building search, summarization, or RAG products, the list deserves a line-by-line read.
Four Cases That Went Wrong
The investigation centers on four query categories. Pancreatic cancer: AI Overviews advised patients to avoid high-fat foods. Anna Jewell of Pancreatic Cancer UK called that “completely incorrect” — these patients need calories, and the bad advice could leave them losing weight ahead of chemotherapy or surgery, jeopardizing their treatment.
Liver blood tests: the “normal range” answer presented a wall of numbers stripped of any context for age, sex, or demographics. British Liver Trust chief executive Pamela Healy warned that asymptomatic patients with serious liver disease could conclude they are healthy and skip follow-up appointments.
Vaginal cancer: the summary listed a pap test as a diagnostic test, which is incorrect. Athena Lamnisos, chief executive of The Eve Appeal, warned that women might dismiss symptoms after a clear cervical screen — and noted that running the same search again produced a different answer each time.
Mental health: summaries for psychosis and eating disorders contained “very dangerous advice” in the words of Stephen Buckley of Mind, potentially discouraging people from seeking help or reinforcing stigmatizing biases.
Google’s Response
A Google spokesperson told the Guardian that many examples were “incomplete screenshots,” that the actual AI Overviews link to reputable sources, that the “vast majority” of Overviews are accurate and on par with the existing featured snippets, and that the company takes action under its policies when context is missed.
The response itself deserves scrutiny. “On par with featured snippets” anchors a new product’s accuracy to a baseline that was already contested. And the “incomplete screenshots” framing concedes the real issue: users judge the summary text in front of them, not the sources behind a click they never make.
A Problem Dating Back to 2024
AI Overviews rolled out to all US users at Google I/O in May 2024. Within three weeks, Google had published a post titled “About last week” defending the feature’s accuracy after the first wave of criticism. By October 2024 it covered more than 100 countries and over a billion users per month; by May 2025, more than 200 countries and territories. The scale kept compounding, while the structural question — can a generated health summary be trusted — has been there since day one.
Sophie Randall of the Patient Information Forum and Stephanie Parker, director of digital at Marie Curie, named the core scenario: “People turn to the internet in moments of worry and crisis.” In that context, the authoritative look of a generated answer is itself a risk amplifier. The Guardian also notes a November study that found AI chatbots giving inaccurate financial advice — a reminder that health is the sharpest edge of a problem that spans every domain where users ask models for consequential guidance.
Lessons for Search and Summary Products
First, non-determinism equals unusable in critical domains. A query that returns a different answer on every run is a novelty in entertainment and a defect in healthcare; caching and answer stabilization belong in baseline requirements. Second, treat provenance as first-class UI. Generated summaries remove the click, and with it the user’s motivation to cross-check; source links cannot be a gray-text footnote. Third, high-risk domains need tiered pipelines. Health, legal, and financial queries should route through more conservative retrieval and generation settings — stronger grounding, less generation freedom — rather than sharing one parameter set with everything else. AI Overviews reaches hundreds of millions of people monthly; its error patterns are free curriculum for anyone building a summary product.
Sources
- Google AI Overviews put people at risk of harm with misleading health advice — The Guardian
- Generative AI in Search: Let Google do the searching for you — Google
- AI Overviews: about last week — Google
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.