AI is getting remarkably good at diagnosing health issues, but doctors are still better at weighing treatment options

Artificial intelligence is increasingly matching or surpassing doctors in diagnostic accuracy. An April 2026 study published in the *New England Journal of Medicine* found OpenAI’s o1 model achieved a 78% accuracy rate in diagnosing complex medical cases, outperforming experienced doctors in emergency room settings. A separate 2024 study showed ChatGPT alone outperformed physicians in diagnosing complicated cases, even when doctors used the AI tool alongside their expertise. While AI excels at pattern recognition—matching symptoms to known conditions—diagnosis represents only half of a doctor’s role. The more challenging task is determining how to treat a patient, which involves weighing multiple factors, including uncertainty and individual patient needs. Experienced doctors rely on mental shortcuts called illness scripts, which help them quickly categorize symptoms and identify deviations that may require further investigation. AI systems like ChatGPT function similarly by predicting patterns from vast datasets, including medical literature. However, treatment decisions often lack a single correct answer, requiring doctors to consider patient history, lifestyle, and other variables. For example, a clear diagnosis like pneumonia may still require tailored treatment based on the patient’s overall health and potential complications. The gap between AI’s diagnostic strength and human expertise in management highlights the need for collaboration rather than replacement. While AI can assist in identifying conditions, doctors remain essential for navigating the complexities of patient care, especially in ambiguous or high-stakes scenarios.