China News Service, Beijing, October 21 (Luo Chunhao) Beanbags are breaking into the clinic and becoming the "third person" between doctors and patients. But most of the time, it is not here to steal doctors' jobs, but to ask questions for patients.
Recently, a "clinic notice" was circulated on the Internet, which said, "Patients who have been diagnosed at Doubao should not seek advice from us again. Please consult Qianwen or Yuanbao instead." Although the picture has been verified by various media outlets and is a spoof by netizens, a fake notice can trigger a lot of discussion, indicating that it hits a real issue: patients walking into the clinic with health information consulted by AI is no longer an isolated phenomenon.
Wang Yue, a professor at the School of Medical Humanities at Peking University, gave an example: A doctor at a large tertiary hospital discovered that a patient was recording secretly with his mobile phone during a consultation. The doctor did not stop him. Instead, he told the patient to change the angle. He could not only record but also video, take it back and listen to it slowly.
Wang Yue said that with the advent of AI, it may become more common for patients to participate more in the diagnosis and treatment process.
In the past, when we discussed AI, we asked how much efficiency it could help doctors improve and whether AI consultation was reliable. Now, the question has changed in another direction: when a patient asks "Why are you different from what he said" with the answer given by AI, how much more time does the doctor have to spend explaining? Where does all this time come from?
Shen Qi, a professor at the Institute of Aging at Fudan University, told Sino-Singapore Health that this year his team conducted a two-end survey, including 75 doctors and 143 middle-aged and elderly residents. The survey results found that about 70% of the doctors surveyed have encountered patients using AI to query health information in outpatient clinics, of which 36% have encountered patients 2-4 times a month, 10.67% 2-3 times a week, and 24% of doctors have seen patients using AI more than 4 times a week.
This leads to a new question: What changes has the "patient-AI-doctor" model brought about?
"In the past, Baidu was used to see a doctor, and later Doubao was diagnosed."
Some netizens joked, "In the past, Baidu was used to diagnose the disease, and later Doubao was diagnosed." The way patients look for information has indeed changed.
Nowadays, advertisements for the health AI assistant "Afu" are overwhelming, and hundreds of millions were spent on related advertising alone in the early days; AI applications such as Doubao, Qianwen, and Yuanbao have also begun to appear on billboards in stations and buildings.
In the past decade, consultations have moved online. In 2015, the first Internet hospital was born, and then Internet medical care and online consultations mushroomed. In 2018, the country issued the "Internet Diagnosis and Treatment Management Measures (Trial)", which clarified that Internet diagnosis and treatment is not allowed for first-time patients.
However, life is not easy for Internet medical platforms: Chunyu Doctor's revenue in 2023 was 101 million yuan, but it dropped to 51.048 million yuan in the first 10 months of 2025, and was eventually "sold"; WeDoctor's cumulative losses exceeded 8 billion yuan from 2021 to the first half of 2024, and the road to profitability is bumpy; Haodafu Online was acquired by Ant Group.
Today, AI is becoming the “first stop” for more and more Chinese people to access health information. QuestMobile data shows that as of March 2026, the number of monthly active users of AI native APPs has reached 440 million.
Shen Qi told Sino-Singapore Health that unlike Baidu search, when people use large AI models, many of them are looking for a "judgment": input age, symptoms, and test results, and it can organize a set of seemingly complete explanations and even tell you what to do next. This interaction can change people's psychology of receiving health information.
But here’s the problem: AI can organize answers based on questions, but it doesn’t mean it knows all the patient’s health conditions. In the survey, 81.33% of the interviewed doctors believed that the most prominent problem with AI health advice is the lack of individualization - the same dizziness and chest tightness may lead to completely different clinical judgments under different ages, underlying diseases and medication conditions. In the past, I was worried that the public would not be able to find reliable information, but now I am even more worried that they will over-believe a piece of seemingly reliable information.
Are hospital doctors driven crazy by beanbags?
Wang Yue believes that the challenge of AI to doctor-patient communication is not a new phenomenon. It emerged with the Internet, but the large model is more intelligent and more like a person than a search engine. In his words, the traditional doctor-centered communication method between doctors and patients is facing “collapse.”
He advocated that doctors should move away from the "paternalistic" doctor-patient relationship and form a friend-like relationship with patients who jointly participate in the diagnosis and treatment process. "Instead of hating learning patients, it is better to cultivate learning patients." Doctors need to help patients screen the information provided by AI and tell them which ones are suitable. In Wang Yue's view, doctors' monopoly on professional knowledge is being broken.
But doctors are not saying "no" to AI en masse. In the above-mentioned survey, 65% of doctors chose "neutral" on patients using AI to query health information, and 16% clearly did not support it.
Shen Qi said that before the survey, the team thought that with AI, patients would be able to check a lot of medical knowledge on their own and would be less dependent on doctors; but the data from residents was not the case: when AI disagrees with doctors, 43% of the respondents would listen directly to the doctor, and 35% would mainly listen to the doctor and hope that the doctor would explain why it is different from the AI. In other words, the doctor’s professional judgment is still the final basis, but patients have higher requirements for “explanation”.
In the future, doctors may have to spend more time communicating, which is a new burden for them, but not necessarily a completely bad thing.
Shen Qi believes that when patients use AI to ask doctors, it’s not because they don’t trust doctors, but because they really want to understand their condition. If doctors can explain the differences clearly, they will have the opportunity to enhance patients' understanding and trust in the diagnosis and treatment plan.
Where do the extra few minutes of consultation time come from?
The problem ultimately comes down to timing. Wang Yue believes that it is necessary to ensure consultation time. He said that many hospitals now give patients about 5 minutes; if they can get 2 more minutes, it will be a qualitative change.
The Lancet published a research report on nearly 28,000 patients in 136 tertiary hospitals in China in 2018. The results show that the consultation time is not simply proportional to patient experience, but there is an obvious "inflection point": within 8 minutes, for every increase in consultation time, the probability of patients giving a higher evaluation increases significantly; after 8 minutes, the increase levels off.
The Hubei Provincial Health Commission has piloted "full consultation" in the "Improving Medical Experience" program from 2023 to 2025: medical institutions at level 2 and above can select some specialists or experts, limit each expert's number sources to about 30 for half a working day, ensure that each patient communicates for an average of more than 6 minutes, and fully explain the patient's condition, interpret test orders, and inform precautions.
Wang Yue believes that to solve the time problem, efforts must be made to standardize doctors’ workload and performance-based compensation system.
As for the role of AI in popularizing health knowledge, preparing for medical treatment, and assisting in chronic disease management, Shen Qi is relatively optimistic. Shen Qi said that for middle-aged and elderly people, the biggest advantage of AI is that it can ask questions repeatedly: when you get the physical examination report, ask the AI first; before going to the hospital, let the AI help sort out symptoms and questions to ask the doctor; when the doctor prescribes medicine, use AI to understand the terminology in the medical order. These scenarios have practical value.
"But when it comes to diagnosis, treatment plans, and medication adjustments, the boundaries must be clear. Especially the elderly, who often suffer from several chronic diseases at the same time and take multiple medications for a long time. Even if AI gives advice that is consistent with general medical knowledge, it may not be suitable for this elderly person."
Shen Qi believes that when evaluating large-scale AI model tools in the future, we should not only look at how many medical questions it can answer and how fast it can answer them, but also whether it can recognize the limitations of its own capabilities and whether it can tell users when they must go to the hospital and what things they cannot handle on their own.
