Microsoft's Diagnostic AI Asked 25 Questions Before Its First Test. Patients Get a Turn Next
The system averaged 25.5 questions before ordering a test, nearly five times the physician average. The published comparisons involved individual generalists working through text-only cases.
Tech · · 2 views
MAI-DxO asked an average of 25.5 questions before ordering its first test. Physicians averaged 5.6. The machine has found a way to make the first appointment feel like the follow-up.
Microsoft says its diagnostic system scored 83.2% on complex New England Journal of Medicine cases, compared with 36.1% for individual generalists using reference tools. On simulated emergency-department cases, it scored 80.3%, against 46.8%. DxO gathers more information, orders tests and revises its diagnosis in stages.
This was a controlled, text-only benchmark. The researchers compared DxO with individual generalists; their paper says it does not establish superiority over specialists, teams or real-world practice. Microsoft funded the study, and all authors are company employees who hold equity.
The next evaluation at Beth Israel Deaconess will involve patients receiving standard care alongside DxO. That brings a new variable to the questionnaire: whether a person still wants to answer after question 25.
Based on: Building Medical Superintelligence, Microsoft AI, October 22, 2026; Microsoft AI diagnostic tool tops generalist physicians, 83% to 36%, Becker's Hospital Review, October 23, 2026; Microsoft's Mustafa Suleyman says its AI outperforms doctors on hard cases as he chases medical superintelligence, TechCrunch, October 22, 2026.