At the 2026 World Internet Conference Wuzhen Summit on November 5, Pingan Group Chief Technology Officer Wang Xiaohang disclosed that the Pingan AI Doctor now handles nearly 300,000 professional services per day, with cumulative usage exceeding 62 million person-times in the first three quarters of 2026, covering pre-diagnosis, in-visit and post-visit care. According to Eastmoney and China Securities Journal, this is Pingan's most systematic disclosure yet of its medical AI operations in complex-disease care.
On capability, the AI Doctor runs a three-tier system: general practice, specialist and MDT (multidisciplinary team) consultation. Two metrics define its complex-disease performance: about 90% consistency between AI-generated MDT plans and expert conclusions, with medical red-line violations held at 0.3%. In April, Pingan's medical large model 3.5 scored 57.27 on the international HealthBench Hard evaluation, ranking first globally at the time.
The AI's role is not to replace clinicians but to run triage and matching: filtering suitable experts by disease stage, clinical evidence and specialist expertise, with professional teams reviewing the match logic and coordinating consultations, registration, tests and admission. Today 25% of demand for Pingan's offline services such as green-channel referrals and accompanied visits is first received and routed by the AI Doctor. Leveraging resources including Peking University Medical Group, the service network spans four layers: online, in-hospital, in-enterprise and at-home.
China's medical-AI narrative is shifting from "diagnostic assistant" to "service network entry point". Insurance and health management give the AI doctor a natural closed loop: high-frequency consultation, matching and follow-up run on AI, while low-frequency, high-value specialist care stays with physicians. A daily volume of 300,000 services shows the former works at scale; the next variable to watch is the share of AI recommendations accepted by physicians in complex cases — the hard metric of medical AI's value conversion.