Google’s AI search feature (AI Overviews) is facing a new trust crisis. A study of more than 50,000 health-related queries found that the AI tool cited YouTube more frequently than any professional medical website when answering medical questions. This finding has raised deep concerns among academics and public health experts about the reliability of AI-generated health information.

The research conducted by SE Ranking mainly analyzed health search data from Germany. The results showed that YouTube ranked first with a 4.43% share in all AI summary sources. In contrast, no hospital network, government health portal, or academic institution came close to this rate. Researchers pointed out that YouTube is not a professional medical publishing platform, and its content includes creators ranging from practicing doctors to health influencers without medical backgrounds, making the rigor of the information difficult to guarantee.

Experts warn that the summaries provided by Google AI in a confident and authoritative tone are subtly influencing the nearly 2 billion users worldwide each month. Hannah van Kolfschooten, a researcher at the University of Basel in Switzerland, noted that this study demonstrates that the risks of AI in the health sector are structural, indicating that Google’s algorithm may prioritize the “visibility” and “popularity” of content rather than its “reliability” in medical terms.

In response, a Google spokesperson stated that AI Overviews aim to showcase high-quality content from reputable sources, and that 96% of the popular YouTube references in the study did come from medical institution channels. However, researchers countered that this only represented a small portion of video samples, with most cited videos still lacking strict medical review. Previously, an investigation by The Guardian also found serious misleading errors in Google AI’s provision of liver function test information.