May 26, 2026
The findings showed that the AI model correctly identified exact or near-correct diagnoses in 67% of emergency room triage cases
Artificial intelligence (AI) is changing discussions about the future of healthcare. A new study from Harvard found that an AI model performed better than human doctors in several emergency room diagnostic tests.
This study, published in the Science Journal and highlighted by AfroTech, looked at how OpenAI’s “o1” large language model performed compared to doctors handling emergency room cases. Researchers from Harvard Medical School, Stanford University, and Beth Israel Deaconess Medical Center took part in this research.
The findings showed that the AI model correctly identified exact or near-correct diagnoses in 67% of emergency room triage cases. In comparison, physicians scored between 50% and 55% under the same conditions. When researchers later provided additional patient information, the AI’s accuracy rose to 82%, still slightly better than that of the doctors.
The researchers evaluated 76 actual emergency room patient cases from a Boston hospital using electronic health records that included demographic information, vital signs, and nurses’ notes. The study found that AI was particularly strong in the early stages of triage, when doctors often must make quick decisions with limited information.
However, researchers and medical experts emphasized that this technology is not meant to replace physicians. A separate report from Vox noted that the AI system analyzed only text-based patient data and could not assess physical symptoms, emotional distress, or body language — factors doctors regularly rely on in emergency settings.
Lead researcher Arjun Manrai described the findings as proof of a significant change happening in medicine as AI tools improve. He told TechCrunch, “We tested the AI model against virtually every benchmark, and it eclipsed both prior models and our physician baselines.”
Experts believe that this technology could eventually act as a second-opinion tool to help doctors enhance diagnostic accuracy and treatment planning, rather than completely replace human care.
RELATED CONTENT: Artificial Intelligence Is Changing The Way Students Pick College Majors
