Google has removed some AI-generated health summaries from its search results following reports that the feature provided inaccurate and potentially harmful medical information.
The summaries, known as AI Overviews, appear at the top of Google search results and are seen by an estimated 2 billion users each month. They use generative artificial intelligence to produce short answers to users’ questions.
Concerns were raised after an investigation published by The Guardian on Saturday, 24 January 2026, reported that AI Overviews gave misleading information about medical tests, including liver function blood tests.
In one example cited by medical experts, Google’s AI Overview provided a single set of “normal” ranges for liver test results without accounting for age, sex, ethnicity, or medical history. Doctors warned that such simplified figures could cause people with early-stage liver disease to believe they were healthy and delay further medical care.
Following the report, Google removed AI Overviews for some health-related searches. A Google spokesperson said the company acted quickly once the issues were identified and maintained that the feature is generally reliable.
However, health organisations, including the British Liver Trust, said that similar inaccurate information could still appear when users rephrased the same medical questions.
Further concerns were raised by a study conducted by SE Ranking, a German search analytics company. The study analysed 50,807 health-related search queries conducted in Germany in December 2025.
The findings showed that AI Overviews appeared in more than 82% of health searches. The study also found that YouTube was the most frequently cited source, accounting for 20,621 citations, or 4.43% of all sources referenced. This was higher than citations from hospital websites, government health portals, or medical associations.
Researchers reported that 66% of cited sources had unverified medical credibility, while less than 1% came from government health agencies or academic journals.
SE Ranking said:
“Most AI-generated health-related answers rely heavily on YouTube, commercial websites, and blogs. This implies that a substantial portion of health answers lack evidence-based support.”
The researchers noted that YouTube is a general video platform where content can be uploaded by anyone, including individuals without medical training.
Google responded by stating that AI Overviews are designed to surface high-quality content and that many health-related YouTube videos are produced by hospitals, clinics, and licensed medical professionals. The company added that the SE Ranking study was limited to German-language searches and could not be applied globally.
Despite this, independent experts expressed concern. Hannah van Kolfschooten, a researcher in AI, health, and law at the University of Basel, said the findings pointed to a deeper problem.
“This study shows that the risks posed by AI Overviews for health are structural, not rare errors,” she said. “Reliance on visibility and popularity rather than medical authority creates real risks for patients.”
