The compound annual growth rate exceeds 29%, and domestic medical large models are accelerating their market penetration.
Introduction
According to the 2026 survey conducted by the American Medical Association (AMA), 81% of the surveyed physicians have used AI in their professional work, while the proportion was only 35% in 2023. Meanwhile, the average number of AI application scenarios used by physicians has also increased from 1.1 to 2.3. AI is increasingly widely integrated into the daily work of physicians.
As physicians' acceptance of AI continues to rise, the intelligent construction of medical institutions has also received more attention.
According to the *2026 Research Report on China's Medical Large Model Industry* released by iResearch, by 2028, the penetration rate of medical large models on the hospital side in the Chinese market will reach 65%, and the proportion of large model-related expenditures in the medical IT construction budget is expected to rise to 20%. From 2025 to 2028, the compound annual growth rates on the hospital, primary care and public health sides all exceed 29%, and the accelerated penetration of medical large models in all scenarios is clear.
From auxiliary tools in the hands of physicians to systematic applications in medical institutions, AI is accelerating its integration into the medical service system. As the scope of application continues to expand, what changes is medical AI undergoing? And how will it further affect the development of the medical industry?
AI Is Becoming a Daily Tool for Physicians
For physicians, the value of AI is increasingly reflected in specific work scenarios.
Diagnosis and treatment work involves a large amount of information processing. From consulting medical literature and referring to clinical guidelines, to organizing medical records, recording diagnosis and treatment processes, and communicating with patients, physicians need to spend a lot of time processing information in addition to clinical judgment. Especially under the background of uneven distribution of medical resources and heavy workload of medical staff, how to improve work efficiency has always been an important issue faced by the medical industry.
The application of AI is providing new solutions for these tasks. According to the survey by the American Medical Association (AMA), summary generation of medical research and diagnosis and treatment standards is the most widely used AI function by physicians, with a usage rate close to 40%; clinical document recording and sorting, patient communication and medical order generation are also important application scenarios. These applications are mainly oriented to specific needs such as information sorting and document processing, providing new tools for physicians to reduce repetitive work and improve information processing efficiency, and gradually integrating AI into daily work processes.
From a broader perspective of medical applications, the capabilities of AI are also extending to professional scenarios such as diagnosis and treatment assistance. The iResearch report points out that medical large models have covered multiple links including intelligent triage, medical record generation, medical image analysis, and clinical decision support. Among them, the application maturity of medical imaging and diagnosis and treatment assistance has reached 75%-85% and 70%-80% respectively, making them relatively mature fields in the development of medical AI.
From Individual Physician Use to Medical Institution Deployment
While physicians are gradually getting familiar with AI tools, medical institutions are also exploring more systematic applications. As application demands become increasingly clear, medical large models are moving from technical tests in partial scenarios to institutional deployment on a larger scale.
The *2026 Research Report on China's Medical Large Model Industry* by iResearch shows that the penetration rate of medical large models on China's hospital side in 2025 is about 30%, and it is expected to rise to 65% by 2028. The penetration rates of primary medical institutions and public health institutions are also expected to grow from 15% and 10% to 50% and 40% respectively. Meanwhile, in the IT construction budget of medical institutions, the proportion of large model-related expenditures is expected to increase from 5% in 2025 to 20% in 2028.
These forecasts reflect that medical institutions' willingness to invest in large model applications is increasing. Hospitals remain the main battlefield for the implementation of medical large models, while primary medical institutions and the public health field show potential for further expansion. With the improvement of technical maturity and the continuous expansion of application scenarios, medical large models are expected to move from pilot projects to a wider range of business applications.
The demands for AI vary across different medical scenarios. Fields such as medical imaging, diagnosis and treatment assistance, and hospital management already have certain application foundations, and large models can participate in clinical information processing, auxiliary decision-making and operation management. In the field of primary healthcare, problems such as relative shortage of high-quality medical resources and heavy workload of professional staff make auxiliary diagnosis and treatment and improvement of medical service efficiency worthy of attention as application directions.
However, the large-scale deployment of medical large models still requires the support of multiple conditions. The report believes that technical maturity, policy environment, commercial closed loop and actual efficiency demands are important factors driving industry penetration. Among them, whether a sustainable business model can be formed and whether the application effect can meet the actual business demands will directly affect the long-term procurement and application of medical institutions.
From Auxiliary Tools to Business Collaboration, Medical AI Moves Towards Deeper Application
The application of medical large models is extending to more complex business scenarios. Early large model applications mainly focused on single-point tasks such as question answering, document generation and information processing. With the development of agent technology, capabilities such as task planning, system collaboration and business process execution have gradually received attention, promoting medical AI to develop towards deeper business collaboration.
This change is closely related to the development of medical AI Agents. Traditional large models mainly generate responses based on user input, while agents, on the basis of the cognitive capabilities of large models, further have capabilities such as task decomposition, tool invocation and continuous feedback. For example, medical AI Agents can connect to systems such as electronic medical records, medical images and medical knowledge bases to assist in tasks including information integration, clinical decision support and medical record generation. In more complex scenarios, multiple professional agents can also provide technical support for applications such as multi-disciplinary consultation and diagnosis-treatment process connection through a collaboration mechanism. These deep collaboration capabilities are still in the stage of continuous exploration and verification.
The expansion of technical capabilities is also changing the competitive focus of the medical large model market. The iResearch report points out that as the performance difference of basic models gradually narrows, it is increasingly difficult to form a sustained competitive advantage by relying solely on model parameters and evaluation scores. Whether the model capabilities can be embedded into real medical scenarios, complete cross-system collaboration and form a business closed loop is becoming an important direction for vendor competition. Even if the same basic large model is adopted, the capabilities of different vendors in engineering implementation and business process integration may bring different application effects.
However, there are still obvious differences in the development of medical AI in different scenarios. The report shows that fields such as medical imaging and diagnosis and treatment assistance already have relatively mature application foundations, while scenarios such as drug R&D and public health are still in the stage of technical verification or early exploration. Medical business has high requirements for accuracy, reliability and data security. In the process of extending agent capabilities to complex scenarios, it is still necessary to continuously verify the application effect in combination with specific business demands.
Conclusion
With the continuous in-depth application of AI in the medical field, the focus of the industry has gradually shifted from technical capabilities to actual value. How to make large models adapt to different medical scenarios, integrate into existing business processes, and form a sustainable business model will continue to affect the future development of medical AI.
Focusing on the application and development of medical AI, the *2026 Research Report on China's Medical Large Model Industry* by iResearch also systematically sorts out the industry driving factors, application scenarios, market competition pattern and commercialization paths, and deeply analyzes the future development trend of medical AI Agents. For more research findings and industry insights, you are welcome to click "Read Original" at the end of the article to get the full report and further understand the development landscape of China's medical large model industry.
This article is from the WeChat Official Account "iResearch" (ID: iresearch-), author: iResearch, published with authorization from 36Kr.