Widespread inaccuracies in AI chatbot health responses pose public health risks, highlighting the urgent need for better oversight and cleaner training data.