ChatGPT Health 'under-triaged' half of medical emergencies in a new study

March 4, 2026 - 04:39

A recent study has cast significant doubt on the reliability of popular AI chatbots for urgent medical advice. When presented with a series of standardized medical scenarios, a leading chatbot consistently downplayed serious health crises. The research found that in over half of the cases where human doctors would have immediately recommended emergency room care, the AI advised that treatment could safely be delayed.

The investigation involved testing the chatbot against recognized medical guidelines for triage, the process of determining the urgency of a patient's condition. The AI's failure to recognize the severity of symptoms presents a clear risk. Patients relying on such platforms for immediate guidance might be incorrectly reassured to stay home when they actually require urgent, life-saving intervention.

Experts emphasize that while artificial intelligence holds promise for administrative tasks and general health information, this study underscores its current limitations in acute care settings. The technology, they caution, lacks the nuanced clinical judgment and ability to interpret complex human symptoms that trained medical professionals provide. The findings serve as a stark warning to the public against using consumer AI chatbots for diagnosing or triaging potentially serious medical conditions, urging individuals to instead contact healthcare providers or emergency services directly when faced with concerning symptoms.