Commentary: Dr ChatGPT is getting remarkably good at diagnosing health problems

Artificial intelligence is rapidly advancing in its ability to diagnose medical conditions, with some AI models now matching or exceeding the accuracy of human physicians in complex cases. A father worried about his toddler's fever and ear pain, and a woman concerned about fatigue and shortness of breath, both turned to AI chatbots for answers. The AI suggested an ear infection for the child and a potential cardiac condition for the woman, providing helpful and likely correct initial assessments. A study from April 2026 found that OpenAI's GPT-4 model achieved a 78% accuracy rate on complex diagnostic cases from the New England Journal of Medicine, outperforming experienced doctors. Another 2024 study indicated that ChatGPT, used independently, surpassed physicians in diagnosing difficult cases, even when physicians had access to the AI themselves. This diagnostic prowess stems from AI's ability to process vast amounts of data and identify patterns, similar to how experienced doctors build 'illness scripts', mental shortcuts based on years of practice that help them categorize symptoms and recognize diseases. However, the article emphasizes that diagnosis is only half the job. The other crucial half is management reasoning: deciding how to treat a patient. While AI can categorize conditions, it currently struggles with the nuanced decision-making required for patient management, especially in cases of uncertainty. Experienced doctors use their illness scripts not only to diagnose but also to notice deviations that might point to less common conditions or require different management strategies. This complex process of prioritizing and managing care, even after a diagnosis is clear, remains a domain where human medical expertise is indispensable.

“

Your child likely has an ear infection.