AI chatbots often outperform doctors in diagnosis but may overprescribe, study finds

A study by Melbourne University found that AI chatbots can often outperform human doctors in diagnosis but pose significant risks like overprescribing and recommending unnecessary tests. "[This] observed tendency of AI...may reflect...lack of real-world accountability mechanisms and potential biases...that reward exhaustive workups," it said. Researchers analysed ERNIE Bot, ChatGPT and DeepSeek.