AI Model Outperforms Doctors in Emergency Triage Diagnoses, Study Finds

May 3, 2026

AI Summary

A Harvard study reveals that OpenAI's AI model diagnosed 67% of emergency patients accurately, surpassing human doctors who achieved 50-55% accuracy. The findings suggest a significant advancement in AI's clinical reasoning capabilities, particularly in high-pressure triage situations.

The study involved a comparison between OpenAI's AI model and human doctors in emergency triage settings at a Boston hospital.
The AI diagnosed patients accurately in 67% of cases, while human doctors achieved 50-55% accuracy.
The AI's accuracy improved to 82% with more detailed information, compared to 70-79% for human doctors, although this difference was not statistically significant.
The AI also outperformed human doctors in creating long-term treatment plans, scoring 89% compared to 34% for doctors using conventional resources.
The study highlights that the AI's performance was based solely on text-based patient data, not on visual or distress signals.
Researchers emphasized that AI is not intended to replace doctors but to assist them in a new care model involving both human and AI collaboration.
Concerns regarding AI errors and accountability in clinical settings were raised, with experts noting the need for a formal framework for AI use in medicine.
The study indicates that AI could serve as a valuable second-opinion tool for clinicians, especially in complex diagnostic scenarios.

openaiemergency medicinediagnosishealthcareai performance

AI Model Outperforms Doctors in Emergency Triage Diagnoses, Study Finds

Related Stories

Leaders Face Challenges in Managing AI Agents in the Workplace

Concerns Rise Among Lawyers Over Use of AI Note-Taking Tools

Emerging 'Wired Belt' highlights political potential of white-collar workers facing AI job risks

Future Workspaces May Embrace Dictation Technology and Whispering