Typing symptoms into Google whenever something feels off has become a reflex for millions of Vietnamese users. Instead of sifting through dozens of familiar blue links, people now get a neat summary generated by artificial intelligence (AI) right at the top. But a startling new study has revealed a troubling truth: the "doctor" behind that medical advice is not a respected hospital or medical journal. The leading source is, of all things, YouTube.
YouTube Is the Number One Source for Google AI
SE Ranking, a platform specializing in search engine optimization (SEO), ran a large-scale analysis of more than 50,000 health-related queries in Berlin in December 2025. It found that Google's AI Overviews feature, the tool that automatically summarizes answers, cited YouTube more often than any other source. Specifically, the video platform appeared in 4.43% of all citations, well ahead of reputable hospital networks, health ministries, and academic websites.
That makes YouTube, the second most visited website in the world, the single most influential factor in shaping how AI explains medical conditions to users. It is an alarming paradox for a tool that Google has always claimed "surfaces high-quality content from reliable sources." In reality, YouTube is not a medical publisher. On an open platform where anyone can post, from leading professors to lifestyle vloggers and even self-styled "online healers," the reliability of information is an extremely unpredictable variable.
In the ranking of sources favored by the AI, YouTube leaves the rest far behind. Germany's public broadcaster NDR comes in second with just over 3%, followed by the respected medical reference site MSD Manuals (2.08%). These figures suggest that even in a market with strict medical oversight like Germany, Google's algorithm is still "dazzled" by the popularity of video rather than the accuracy of medical text.
When Algorithms Value Virality Over Accuracy
Responding to the findings, a Google representative said its AI systems draw on reliable information "regardless of format," and stressed that many medical professionals also run their own YouTube channels. The search giant argued that among the 25 most-cited YouTube videos, as many as 96% come from verified medical channels. The researchers quickly pointed out the flaw in that argument, however: those 25 videos account for less than 1% of all YouTube citations in the study, far too small a share to represent the full picture.
The core problem lies in how the AI works. Hannah van Kolfschooten, a researcher at the University of Basel, says that relying so heavily on YouTube shows the system is being driven by "visibility and popularity" rather than "medical reliability." In other words, a folk-remedy video with millions of views and heavy engagement could be ranked by the AI as more "important" than a dry but accurate scientific report from a frontline hospital.
These concerns are not unfounded. Earlier, an investigation by The Guardian documented multiple cases in which Google's AI produced misleading and even dangerous medical summaries, including false information about liver function tests. Although Google temporarily pulled the AI from some sensitive queries, the SE Ranking study suggests this is a systemic flaw rather than a random glitch. When AI learns from popularity, misleading but appealing information will always have a chance to creep into your health advice.
In short, while Google works on refining its algorithm, users need to keep a cool head. AI Overviews can be handy for getting a rough idea, but it should never be treated as a final diagnosis. When your health is at stake, check the information against reliable sources or consult a doctor, rather than entrusting your life to algorithms that seem to be "addicted" to watching YouTube.
