The use of AI chatbots has surged recently, with users seeking everything from vacation ideas to medical advice. These chatbots can be useful for general inquiries, but they are not reliable sources for medical guidance. A recent study has shown that half of the medical advice provided by AI chatbots is inaccurate or misleading.
A Closer Look at the Study
Researchers from the US, Canada, and the UK tested five popular AI chatbots, ChatGPT, Gemini, Meta AI, Grok, and DeepSeek. They asked 10 questions in five different categories to each chatbot. Shockingly, 50% of the responses were problematic, and 20% were highly problematic. These answers were unreliable, making them unsuitable for health-related inquiries.
The study found that these chatbots performed relatively well with closed-ended questions and queries about vaccines or cancer. However, their answers to open-ended questions, stem cell research, and nutrition were poor and unreliable.
The Absence of References
Another alarming finding from the study was that these chatbots provided answers with complete confidence, yet none offered accurate or comprehensive reference lists for their advice. Only Meta AI refused to answer two questions, while the other chatbots answered all questions. This lack of transparency is concerning, as it shows that these chatbots are not equipped to provide valid medical advice.
The Growing Trend of AI Health Tools
Despite the study’s findings, the popularity of AI chatbots for medical advice continues to rise. OpenAI reported that over 200 million people ask ChatGPT health and wellness-related questions each week. Recognizing this demand, OpenAI integrated health tools into ChatGPT in January. Following this, companies like Anthropic and Microsoft have also included health tools in their chatbots. However, these AI tools are still not designed to diagnose diseases or offer reliable medical recommendations.
