Half of AI Health Advice Is Wrong—And Seems Just Right

A peer-reviewed audit in BMJ Open found that nearly 50% of health responses from five major AI chatbots were problematic, with fabricated sources and confident delivery.