Google AI search summaries give cancer patients wrong advice

Google's "AI Overviews" feature provided inaccurate and dangerous medical advice to cancer and liver disease patients, potentially causing physical harm by deterring them from seeking life-saving treatment or providing false reassurance about abnormal test results. What happened Between late 2024 and early 2026, Google’s generative AI summaries served misleading health information to users in the UK and globally. An investigation by The Guardian in January 2026 revealed that the system hallucinated "normal" ranges for liver function tests and provided incorrect treatment guidance for cancer. Experts warned that the AI's confident tone could lead seriously ill patients to believe they were healthy, causing them to skip essential medical appointments. Why it happened The fracas was driven by "hallucinations" common in Large Language Models (LLMs), where the system prioritises generating a fluent response over factual accuracy. Transparency limitations contributed significantly: Google initially placed health disclaimers in small, light fonts that only appeared after a user clicked "Show more," meaning the most dangerous advice was presented without immediate warning. Furthermore, the system often prioritised Reddit or unverified social media content over peer-reviewed medical data. What it means For patients, this creates a high risk of "digital health inequality," where those relying on search engines for quick answers receive life-threatening misinformation. For society, it erodes trust in established health authorities and pressures the medical system to correct widespread "AI-generated myths." For policymakers, it highlights the urgent need for strict regulation of AI in "Your Money Your Life" (YMYL) categories, suggesting that medical AI should be regulated as a medical device rather than a search feature. System 🤖 AI Overviews Developer: Google Country: Multiple Sector: Health Purpose: Answer health a nd medical questions Technology: Generative AI Issue: Accuracy/reliability; Anthropomorphism; Consent; Safety; Transparency Timeline ⏰ May 2024. Google launches AI Overviews broadly following Google I/O. Late 2024. Rapid expansion of AI summaries into clinical categories (symptoms/treatments). January 2, 2026. The Guardian publishes an investigation exposing "dangerous" medical inaccuracies in AI Overviews. January 11, 2026. Google begins removing AI Overviews for specific high-risk queries (e.g., liver function tests) following the investigation. February 2026. Health charities and the UK Patient Information Forum (PIF) launch a joint report on AI search risks. March 2026. Google scraps a related AI search feature that crowdsourced amateur medical advice.

Date it happened

2024-01-01

Product, system or model

AI Overviews

This is a record of what was reported, not a finding that anyone broke the law. If it names your organisation and you believe it is wrong, the corrections process is free and open to everyone.