AI chatbots fail medical misinformation test, returning inaccurate and fabricated advice

Summary: It might be tempting to ask an AI chatbot for health advice, but you should probably hold off. A recent study tested five popular AI chatbots (including ChatGPT, Gemini, and Grok) with 50 common health questions about things like cancer, vaccines, and nutrition.

The results? Almost half of the AI's answers were problematic. Even worse, the chatbots sounded incredibly confident while giving bad advice, and they often made up fake sources to back up their claims. They struggled the most with questions about nutrition and stem cells. Plus, the answers were written at a college reading level, making them hard for the average person to understand. Until these systems improve, it is best to leave the medical advice to real doctors.

Source: psypost URL: psypost.org/ai-chatbots-fail-medical-misinformation-test-returning-inaccurate-and-fabricated-advice/ (opens in new tab)