A collection of 2026 studies showing AI matching or outperforming ER physicians — from DeepMind's AI co-clinician to the Harvard-Science paper on OpenAI o1 diagnosing 67% of ER patients correctly.
Google DeepMind introduced the AI co-clinician, a multimodal AI agent research initiative to support healthcare workers. In blind evaluations, the system achieved zero critical errors in 97 of 98 primary care queries and matched physician-level performance on medication knowledge benchmarks.
A Harvard Medical School trial found OpenAI's o1 reasoning model correctly diagnosed 67% of ER patients, beating human physicians at 50-55%. With more clinical data, o1 reached 82% accuracy and dominated long-term treatment planning at 89% versus doctors' 34%.
A peer-reviewed study published in Science tested OpenAI's o1 on 76 real ER triage cases and found it achieved exact or near-exact diagnoses 67% of the time, versus 55% and 50% for two attending physicians who received identical patient data.
A new study published in Science found that a state-of-the-art LLM matched or exceeded human emergency physicians in diagnostic choices, emergency triage, and next-step management decisions using real ER data and hundreds of physician comparisons. Researchers say the results call for collaborative care models, not AI replacement of doctors.
The important medical AI story here is not replacement but reliability. Google DeepMind says its AI co-clinician produced zero critical errors in 97 of 98 realistic primary-care queries, while physicians still beat it overall in multimodal telemedicine simulations.