More and more people are turning to AI-based chatbots, such as ChatGPT, for health advice amid increasing pressure on healthcare systems and long waits to see doctors. According to a survey, about one in six adults in the US uses such tools at least once a month for self-assessment of their health.
A study conducted by a group of Oxford researchers showed that users of chatbots often do not achieve better results than those who use regular internet searches or rely on their own knowledge. The study involved more than 1,300 people who were presented with medical scenarios created by doctors. Participants had to identify possible diseases and look for solutions using chatbots, including GPT-4o, Command R+ from Cohere, and Llama 3 from Meta, as well as other available tools.
It turned out that AI model responses often contained both helpful and questionable recommendations, and participants frequently missed important details when formulating their queries. This led to users being less likely to correctly identify diseases and underestimating the seriousness of certain conditions.
Amid the active implementation of AI in healthcare, major companies such as Apple, Amazon, and Microsoft are developing new tools for data analysis and patient support. However, both the medical community and AI developers emphasize that such systems still require thorough testing in real-world conditions and should not serve as the primary source of health decisions. Experts recommend consulting trusted sources of information and professionals when making treatment-related decisions.
