Nearly two billion people each month interact with AI Overviews, an AI-powered search feature from Google that provides summaries of users’ queries.
A new study has revealed a troubling pattern in some of these responses: when users ask health-related questions, AI Overviews appears to rely on YouTube significantly more than on trusted medical websites.
Since its launch, AI Overviews has faced its share of controversy, ranging from early reports suggesting the product delivered nonsensical answers, to a wave of lawsuits filed by publishing companies and groups claiming the feature harms organic web traffic.
The latest concerns emerged from an investigation by The Guardian, published on January 2, which alleged that the tool tends to provide incorrect, misleading, and potentially dangerous health guidance. Google denied those claims at the time.
A new study conducted by the AI-driven SEO platform SE Ranking, published on January 14, found that AI Overviews cites YouTube videos two to three times more often than trusted medical websites when users search for health information. Google, however, insists that this does not tell the full story.
From the perspective of artificial intelligence, all content exists within a single source.
To understand how the feature gathers health guidance online, researchers at SE Ranking analyzed more than 50,000 health-related Google searches conducted by German users. Germany was selected, according to the study’s authors, because of its strict healthcare system.
The researchers wrote in their report:
“If AI systems rely heavily on non-medical or unreliable sources even in such an environment, this suggests that the issue may extend beyond any single country.”
SE Ranking found that only 34 percent of AI Overview results came from trusted medical sources—defined as medical and governmental institutions, academic journals, and similar outlets—while the remaining 66 percent originated from general or non-specialized sources, such as commercial websites or blogs.