A herder in Balikun county used to drive for hours to see a specialist. Now he sits at home, turns on his television, and talks to a doctor through a video link, with an electronic prescription and a remote image reading flowing through the same connection. In a different county, an emergency drug order that once took at least ninety minutes to reach the farthest township clinic now arrives in twenty-five — by drone.
These are not demonstrations staged for a press conference. They are pieces of a programme that Xinjiang has been building since its health authority and three other departments jointly issued an implementation plan for "AI + medical and health" covering 2026 to 2028. The plan was officially issued, not merely drafted, and it sets a concrete target: by 2028, AI-supported diagnosis and image recognition should be in routine use at township health centres (乡镇卫生院) and community health service centres across the region.
A "cloud–network–end" consultation platform
China Telecom's Xinjiang branch has built what it calls a cloud–network–end integrated video consultation platform, and the company says it is already deployed in Hami city. A herder in the Kazakh autonomous county of Balikun can consult a doctor over television, with electronic prescription circulation, remote imaging diagnosis and data from wearable devices all available in the same session.
- The platform is described as connecting a central hospital's capability to the "last mile" of a sparsely populated region.
- Xinjiang is vast — larger than many countries — and specialists are concentrated in cities, so distance, not just cost, is the barrier.
- According to the regional health commission's own explanation of the policy, AI cannot replace doctors; the aim is to lower the risk of missed diagnosis and misjudgement by grassroots clinicians.
The design choices are deliberately unglamorous. Instead of a new app to download, the interface is a television, a device already in nearly every home. The regional digital development bureau framed the value of AI as carrying a central hospital's capability to the periphery, according to a summary carried by Xinjiang Daily in November.
The drone line
In Shufu county, in the south of the region, a drone route for medical supplies was opened — described by local media as the first such line in southern Xinjiang. The reported effect is a drop in average emergency response time of about 60 percent. That average should not be confused with the longest single leg: moving emergency drugs from the county hospital to the farthest township clinic took at least one and a half hours before the route; now it takes about twenty-five minutes — a cut of roughly 70 percent on that one leg, not a region-wide figure.
According to Li An, chief scientist at BrainNet (脑机网), the hard part of rural AI deployment is rarely the model — it is the last leg, where a diagnosis has to become a physical action before a patient's condition changes. The drone line is a literal example of that last leg.
What the policy actually commits to
It is worth separating the plan from the projects. The 2026–2028 plan, issued by the regional health commission, development and reform commission, industry and information technology department and digital development bureau, is a policy document with targets. The Hami video platform and the Shufu drone line are operating projects that may or may not have been created by that document. Chinese official coverage presents them together, but the causal link is not spelled out.
The plan's stated ambitions extend further:
- Build a unified "AI + medical and health" database and health-data standards.
- Pilot smart hospitals and scale up AI-supported diagnosis, medical-image analysis and clinical decision support.
- Push intelligent assistance in diagnosis and image recognition into routine grassroots use by 2028.
- Use a "general model + specialty model + scenario agent" matrix covering clinical care, public health, chronic disease management and traditional Chinese medicine (中医).
Some of these are already visible elsewhere. Nationally, more than 8,000 primary medical institutions have connected to cloud AI imaging services, according to a National Development and Reform Commission briefing in early November — a separate programme that shows the direction of travel.
Honest limitations
The 60 percent response-time improvement and the 25-minute delivery are figures reported by regional media and tied to a specific county route; they are not a measured average across Xinjiang. The video platform's capabilities are described by the operator and local press, not by an independent clinical evaluation, and no data on diagnostic accuracy or patient outcomes is provided here. "Routine use by 2028" is a policy target, not an accomplished fact. The article does not cover data-privacy safeguards for video consultations or the cost of the drone service, both of which matter for whether the model is sustainable. Finally, Xinjiang's geography and concentrated funding make this a demanding case to replicate elsewhere.
What readers can do now
- If you work in rural health, ask whether a proposed AI tool fits an existing device in the home or clinic. The television interface succeeded precisely because it did not require a new gadget.
- Treat "AI-supported diagnosis" as a second opinion, not a replacement. The regional health commission itself says the technology cannot substitute for doctors.
- For anyone evaluating a similar drone or telemedicine programme, demand the full route data — number of deliveries, failures, and cost per delivery — rather than a single headline percentage.
