Study evaluates medical advice from AI -chatbots and other sources

A team of AI and medical researchers, affiliated with various institutions in the UK and the US, has tested the accuracy of medical information and advice from LLMS to users. In their paper posted on the Arxiv Preprint server, the group describes how they asked 1,298 volunteers to ask chatbots for medical advice. They then compared the results with advice from other online sources or the common sense of the user.

Making a journey to see a doctor for an ailment can be time-consuming, embarrassing and stressful and sometimes expensive. That is why people started looking at their local chatbot in many places, such as Chatgpt, for advice. In this new effort, the researchers wanted to know how good that advice could be.

Earlier research has shown that AI apps can achieve almost perfect scores on exams for medical license and also perform very well on other medical benchmarks. But to date, little work has been done to see how good such skills translate into the field. Earlier research has also shown that a lot of skill and experience is needed for doctors to ask their patients better questions and/or to give better answers to their questions.

To test the accuracy of medical advice from LLMS, the team compared their advice with other sources. They asked 1,298 randomly assigned volunteers to use an AI-Chatbot (such as Command R+, Llama 3 or GPT-4O) or to use all the sources they would normally consult at home-such as internet search assignments or their own knowledge while they are confronted with a medical situation. The researchers then compared the accuracy of the advice they have given through the chatbots with that of the control group.

All conversations between the volunteers and chatbots were recorded and sent to the research team for evaluation. The researchers discovered that the volunteers often left relevant information during their questions, making it more difficult for the chatbot to gain a complete insight into the ailment. The result, the team suggests, was many breakdowns of two directions.

When comparing possible causes of a disease and treatment options that are proposed by the chatbots with other sources – such as other online medical sites – and even the volunteer’s own intuition, the researchers found the advice given by the chatbots in some circumstances and worse in others. Rarely did they find any proof that the LLMS offers better advice.

They also found many examples where the use of a chatbot was less inclined to identify their ailment correctly and to underestimate the severity of their problem. They conclude by suggesting that people use a more trusted source of information when searching for medical advice.

© 2025 Science X Network