You walk into a hospital you have never visited, in a city you do not know. The signs are a blur, and the queue at the information desk is long. Your mother cannot explain her symptoms clearly in the local dialect.
On your phone, a digital voice already knows which department to send you to, which doctor has an open slot, and what last month's lab report actually means. In China's Zhejiang province, that voice is not a future demo. It is a service called AnZhenEr (安诊儿), and tens of millions of residents already use it.
Key takeaways
- AnZhenEr (安诊儿) is a province-wide AI health companion built by the Zhejiang Provincial Health Commission together with Ant Group (蚂蚁集团), first launched in November 2023 and now in its 3.0 iteration.
- It runs on Ant's medical large model (医疗大模型), Bailing (百灵), and links more than 2,000 medical institutions across all 11 prefecture-level cities in Zhejiang.
- Reported reach is large: over 20 million users and more than 200 million cumulative service interactions, with an average saving of about 20 minutes per visit.
- It was showcased at the 79th World Health Assembly in Geneva in May 2026 as a case study in equitable access to AI health services.
- The same playbook has been extended nationwide through a separate app, "Ant A Fu" (蚂蚁阿福), which reports over 100 million users.
What the service actually does
AnZhenEr (安诊儿) is better described as a digital health human (数字健康人) than a chatbot. A patient opens it through Alipay, a dedicated app, or the provincial government service portal, then talks to it in natural language.
The system handles the full arc of a visit. Before the appointment, it triages symptoms and matches them to an available specialist. During the visit, a "cloud accompaniment" feature pushes step-by-step guidance — where to pay, which floor, when the result is ready. Afterward, it interprets test reports and can surface reminders for follow-up. For chronic and elderly patients, the most cited benefit is simply not getting lost in a large hospital.
One widely reported feature is the "doctor digital twin" (名医数字孪生): a clone trained on a specific physician's case history and reasoning style, available for round-the-clock Q&A on that specialty. Another is the family-doctor assistant, which helps rural users reach primary-care resources without traveling to a city hospital.
The model behind the companion
Bailing (百灵) is Ant's own large model (大模型) trained on medical text, and AnZhenEr (安诊儿) is one of its first public-facing deployments. Unlike a general assistant, it is tuned for healthcare dialogue and wired into appointment, payment and record systems. The 2.0 upgrade in December 2024 added more than 30 online and offline medical scenarios and a new multimodal interface, letting users tap a phone against a labeled ring in some hospitals to summon the assistant and settle bills through medical insurance, commercial insurance or out-of-pocket payment in one step.
Why a province, not a startup, built this
China's public hospitals are crowded, and specialist access is uneven between big cities and counties. Zhejiang has spent years building a provincial health-data backbone, which gave AnZhenEr (安诊儿) something most AI health tools lack: a real connection to hospital systems, not just a conversational front end.
The economic logic is mundane but powerful. If an AI companion shaves 20 minutes off every visit for millions of people, the cumulative time saved for patients and the load reduced on reception desks is substantial. Ant Group has said some hospitals saw front-desk inquiry volume drop by roughly half after deployment, freeing staff for tasks that need a human.
For elderly patients who struggle with hospital apps, the spoken, guided experience removes a barrier that pure self-service portals never solved. For rural patients, the doctor digital twin extends a city specialist's reach without a three-hour bus ride.
According to Li An, chief scientist at BrainNet (脑机网), the more interesting move is not the chatbot itself but the shift from "search a hospital website" to "a continuous companion that already knows your records" — a pattern other public services are likely to copy.
From one province to the country
The national version, "Ant A Fu" (蚂蚁阿福), reports more than 100 million users and answers over 10 million health questions a day. Notably, about 55% of its users come from cities below tier-three, suggesting the tool reaches exactly the populations with the weakest local medical resources.
At the Geneva assembly, Chinese officials presented AnZhenEr (安诊儿) as a replicable template: a public agency sets the clinical and safety rules, a technology partner supplies the model and the distribution, and the result is a shared utility rather than a paid add-on. That framing matters because it positions AI health assistance as public infrastructure, not a premium feature.
Honest limitations
The numbers above come from Chinese state media and Ant Group's own announcements, not from an independent clinical audit. "Satisfaction rate" and "time saved" are operator-reported and should be read as indicative, not peer-reviewed.
AnZhenEr (安诊儿) supports and routes care; it does not replace diagnosis by a licensed physician, and its advice can still be wrong or incomplete. Coverage is concentrated in Zhejiang, and the depth of service varies by hospital. Privacy is another open question: a companion this intimate depends on access to personal health records, and the safeguards around consent and data use are described by the operator rather than verified by outside reviewers.
What readers can do now
- If you or family live in Zhejiang, open Alipay and search "安诊儿" to test triage and report-reading before your next hospital trip.
- Treat any AI health companion as a navigator, not a doctor — bring its summaries to a real clinician and confirm before acting.
- Watch whether your own region copies the model; the Geneva showcase signals this is now an exportable public-service template, not a one-off pilot.