The news: Google removed AI Overview responses to a few health questions after an investigation by The Guardian found that some summaries contained misleading information.

Digging into the details: Earlier this month, a report from The Guardian flagged examples of AI-generated summaries at the top of Google Search results for misleading health guidance.

For instance:

• A search for “vaginal cancer symptoms and tests” falsely listed a Pap test as a vaginal cancer exam.
• The queries “what is the normal range for liver blood tests” and “what is the normal range for liver function tests” returned broad responses that failed to account for patient differences such as age, sex, ethnicity, or nationality.
• Another AI Overview advised people with pancreatic cancer to avoid high-fat foods—guidance that isn’t wrong but lacks crucial context, as it could lead patients to consume too few calories while undergoing cancer treatment.

Google has since disabled AI Overviews for queries about liver tests, per The Guardian. The search giant said AI Overviews are as accurate as other search features, such as featured snippets, and that it takes action when the tech misinterprets online content.

Why it matters: AI Overviews are regularly showing up as responses to people’s search engine health queries.

• Among people who use search engines for questions about a medical symptom or condition, 65% say they have seen AI-generated responses at the top of the results, according to a July 2025 survey from the Annenberg Public Policy Center (APPC) of the University of Pennsylvania.
• 89% of healthcare keywords now generate an AI Overview, per December 2025 data from BrightEdge. Health searches often trigger AI Overviews, which favor the question-based queries typical of health-related topics (e.g., “what is normal blood pressure” or “how to lower cholesterol”).
• Still, most respondents to Annenberg’s survey view AI Overviews for health questions as only somewhat reliable (55%), with just 8% rating them very reliable.

Implications for AI companies: The Guardian’s findings indicate that AI responses to some health queries may omit important context, rather than be wholly unreliable. Still, Google doesn’t want any perception that its LLMs deliver faulty health guidance, especially when competitors OpenAI and Anthropic just released consumer health AI tools.

AI players that encourage consumers to use their platforms for faster health insights must avoid overstating their technology’s reliability while models mature and be transparent about how they collaborate with medical experts to test and refine their AI.