After a Guardian probe found Google’s AI Overviews giving dangerous health advice on cancer, liver tests, and mental health, Google pulled some results but insists most answers are accurate and that it corrects or removes harmful summaries.
In January 2026, Google withdrew a number of medically focused AI Overview responses from its search pages after a Guardian investigation revealed that the tool had been surfacing hazardous false health information. The coverage detailed multiple situations in which generative AI summaries produced incorrect medical guidance that specialists characterized as "alarming" and "potentially life-threatening".
Key Misinformation Reported
The investigation identified several concrete examples of unsafe medical recommendations:
-
Pancreatic Cancer: For people with pancreatic cancer, the AI system incorrectly suggested steering clear of high-fat foods, even though clinicians emphasize that patients typically need a high-calorie intake; following the AI’s advice could undermine their strength and ability to undergo surgery or other treatments.
-
Liver Function Tests: In answering questions about liver blood test results, the AI summaries offered misleading “normal” value ranges that did not adjust for factors such as age, sex, or ethnicity, raising the risk that patients with serious liver disease might wrongly conclude that their lab results were fine.
-
Women’s Cancer Screening: One AI-generated overview mistakenly described Pap tests as a routine screening method for vaginal cancer, when they are in fact used to screen for cervical cancer.
-
Mental Health: For mental health topics including psychosis and eating disorders, the tool reportedly produced inaccurate and potentially damaging advice that might discourage people from seeking qualified professional care.
Google’s Response and Removals
Action Taken
Google turned off AI Overviews for certain health-related searches, among them queries such as "what is the normal range for liver blood tests" and "what is the normal range for liver function tests".
Defense
A company representative argued that the overwhelming majority of AI Overview responses are correct and point users to trustworthy references, and further contended that many of the Guardian’s highlighted examples relied on "incomplete screenshots" that did not show the full context.
Ongoing Policy
Google continues to assert that it intervenes under its existing policies whenever the AI system misreads source material or omits crucial context, and that it adjusts or removes results in such cases.