Google DeepMind AI Co-Clinician Matches or Beats Primary Care Physicians in 68 of 140 Clinical Areas in Randomized Study

Google DeepMind published research on its 'AI co-clinician' system in early May 2026, reporting results from a randomized crossover simulation involving 120 hypothetical telemedical encounters. The system performed comparably to or better than primary care physicians in 68 of 140 assessed areas and recorded zero critical errors in 97 of 98 realistic primary care queries. DeepMind frames the model as part of a 'triadic care' model — AI serves as an assistant to both patients and clinicians, not a replacement for physician judgment. The research is explicitly positioned as addressing the WHO-projected shortfall of 10 million health workers by 2030: in a world where primary care access is rationed by physician supply, an AI that reliably handles the majority of routine primary care queries could dramatically extend the effective reach of each physician. Unlike previous AI diagnostic benchmarks that focused on structured test cases or multiple-choice medical board questions, DeepMind's simulation used realistic, unstructured clinical encounters — making the results more representative of actual telemedical deployment conditions. The zero critical-error rate in 97 of 98 queries is the most clinically significant result: the threshold for clinical deployment is not 'as good as a physician on average' but 'reliable enough not to cause serious harm.' The research joins Harvard Medical School's May 2026 Science publication showing OpenAI o1 outperforming ER physicians in diagnostic accuracy as evidence that AI clinical performance is approaching deployment-ready thresholds.