Study Finds Mixed Results in AI Chatbots' Medical Advice Accuracy

A recent study conducted by researchers from institutions in the UK and the US has evaluated the reliability of medical information provided by large language models (LLMs) such as ChatGPT, Llama 3, and Command R+. The research involved 1,298 volunteers who posed medical queries to these chatbots or relied on their own knowledge and online resources. The team analyzed how the advice from chatbots compared to reputable online medical sources and the volunteers' personal judgments.

The study revealed that many users tend to omit critical details in their queries, which hampers the chatbot's ability to deliver accurate advice due to communication gaps. When the researchers compared the suggested causes of ailments and treatment options, they found that chatbot advice was often comparable to or less accurate than other sources, rarely surpassing them.

Importantly, the findings indicated that reliance on chatbots could sometimes lead individuals to underestimate the severity of their health issues or misidentify their conditions. The researchers recommend that people seek information from trusted and verified healthcare sources rather than solely depending on AI chatbots for medical guidance.

This research underscores the current limitations of AI-driven medical advice, emphasizing the importance of consulting healthcare professionals for accurate diagnosis and treatment. The findings highlight the necessity for improving AI communication skills and ensuring user awareness of the technology's boundaries.

For more details, visit the original study: arXiv:2504.18919.