A resident in Hangzhou, Zhejiang, developed a red rash on their arm; after taking a photo and uploading it to an AI application, they received identification results and care recommendations within two minutes. In Beijing, a young couple was woken at dawn by their child's fever. They first consulted an AI, which advised them to monitor the child at home based on triage guidelines. By morning, the child's fever had subsided, sparing them a trip to the emergency department. An elderly person living alone in Sichuan inputs their daily blood pressure and blood glucose readings into an AI-powered chronic disease management assistant. With a single query, they can instantly learn whether their values are within the safe range, receive dietary adjustment recommendations, and determine whether to contact a doctor.
Three stories, one common feature: AI has become the "first point of contact" for diagnosis. In the past, patients had to follow doctors, sign up, line up and wait. Nowadays, AI is becoming a "pre-clinic" for many people seeking medical care.
Doctor-patient consultations are beginning to change.
The number of consultation scenarios has increased -
From the clinic to the fingertips
In June this year, a health AI application upgraded its "skin photo" feature, expanding the range of identifiable skin conditions from 50 to over 100 types, covering 99% of common skin issues encountered in online medical consultations.At the same time, the app launched a "doctor verification" service—after receiving an AI-generated answer, users can choose to invite a physician from a top-tier (Grade 3A) hospital to review the result.Test data shows that 15% of users opt for "physician review," with an agreement rate exceeding 90% between AI analysis results and physician assessments.
The application of AI-driven follow-up inquiries is also rapidly expanding at the grassroots level. Zhang Shirong, a 73-year-old resident of Yuanhe Street in Xiangcheng District, Suzhou, Jiangsu, suffers from chronic obstructive pulmonary disease, When symptoms worsen, first consult the "AI Personal Health Assistant" on the local health service platform. A response quickly appears on the screen: "Considering your history of chronic obstructive pulmonary disease (COPD), the increased coughing is likely related to exposure to cold air. You may observe for now..." The following day, when he visited the community health service center, his medical history was precisely retrieved by the "AI General Practitioner," and a standardized electronic outpatient medical record was automatically generated.
Liu Zhenlong, associate chief physician at the Department of Sports Medicine, Peking University Third Hospital, offered a fundamental assessment: "The exploration of the 'AI + Doctor' human-machine collaboration model will further drive the integration and innovation of AI technology with professional medical services." He stated: "The application of AI in the healthcare sector is not intended to replace doctors, but to assist them in better addressing the public's needs for medical care and treatment."
But the convenience of the scene also brings new risks. In August this year, the case of a 3-year-old boy in Zhejiang, Yang Yang (not his real name), was sobering. Yangyang has a history of asthma. His mother entered keywords such as "allergy," "asthma," "cough," and "nocturnal wheezing" into an AI system, which promptly provided a diagnosis of "acute exacerbation of bronchial asthma." Yang Yang mother accordingly refused the doctor's advice chest X-ray plain film examination, insist on only prescribing atomization medicine home. Two days later, the child's condition worsened and his breathing was short. When he returned to the hospital overnight, he found a large area of inflammation in both lungs, and he was diagnosed as a bacterial infection with pulmonary virus pneumonia. He was urgently hospitalized. "If they had listened to me earlier, it wouldn't have been so serious." Li Lan, director of pediatrics at Zhejiang Provincial Medical Hospital, said reluctantly.
The questioning scene moved from the clinic to the fingertip, and the threshold was lowered. But while AI can answer "What's wrong with me?", it cannot take responsibility for that answer.
The main body of the application expanded -
From the patient to the doctor
The wave of AI-powered medical consultations is sweeping across the landscape at an astonishing pace. According to statistics, a certain AI assistant handles between 20 to 30 million health-related consultations per day. According to data from the National Health Commission, from January to November 2025, the average daily number of visits to hospitals nationwide will be about 12.69 million. The volume of AI-based medical consultations has surpassed the total daily outpatient capacity of all hospitals across the country.
Patients are no longer merely waiting for doctors to provide answers; they enter the clinic armed with insights from AI, engaging in dialogue with their physicians.
The experience of Zhou Jun, a 32-year-old resident of Wuhan, Hubei Province, is quite representative. His mother was admitted to the respiratory intensive care unit at a local hospital due to physical discomfort. That evening, upon returning home, he uploaded her lab results to an AI platform. "I initially just wanted to understand what those medical terms meant, but unexpectedly, the AI directly informed me that a certain value had already reached a critical level." Zhou Jun promptly consulted several doctors online, and only after discussing the matter did he confirm his condition. "If I hadn't asked the AI, I might have missed this entirely," he said.
Qin Yan, an emergency doctor at a Shanghai hospital, has a deeper view of this phenomenon:“Previously, patients had to be told 'what's wrong with you,' "What to test for and how to treat" now takes the step of explaining why "my judgment is different from what you found" - which should become the new normal of doctor-patient communication.”However, she also acknowledged that if patients rely heavily on AI-generated information and the doctor's opinion conflicts with it, trust between the patient and the physician can easily erode—"this represents a deeper underlying risk."
Doctors are also beginning to use AI themselves. In 2026, Dr. Duan Tao, Director of the Obstetrics and Gynecology Department at Tongji University Affiliated Oriental Hospital, saw his "AI avatar" answer over 700,000 questions for 160,000 patients within six months, while he himself typically sees only 20 patients per day. At the Dongping Street Health Center in Lishui District, Nanjing, Jiangsu Province, an AI-assisted consultation system has been integrated into departments such as general practice, emergency care, and obstetrics and gynecology. When doctors visit, the system automatically enters the medical records, indicates the direction of the diagnosis, and recommends examination items. The errors in the medical reports are significantly reduced, and the efficiency of the visit is increased by about 30%. The hospital's officials said that doctors "up their heads" from the paper and can speak more with patients, reducing the risk of errors. However, they also found that diagnostic accuracy varies across different AI software: "AI can reduce workload, but it cannot replace doctors' signatures."
The application of AI in diagnosis and treatment yields more significant results in the field of rare diseases. Liu Zhenlong told the reporter that he once encountered a child suffering from recurrent knee pain. The parents took the child to several hospitals, but were repeatedly advised to "rest more." Dr. Liu ordered an MRI for the child, and AI-assisted image analysis indicated a rare diagnosis of "osteochondritis dissecans." He also treated a patient with recurrent shoulder dislocations and abnormal skin elasticity; AI suggested screening for hereditary collagen metabolism disorders, leading to a final diagnosis of Ehlers-Danlos syndrome. "For rare and complex conditions that these surgeons encounter infrequently, AI can provide valuable references, helping us broaden our perspectives and avoid unnecessary or blind surgical interventions." Liu Zhenlong said.
Thus, we must also ask: when both patients and doctors use AI, where does the boundary of this relationship lie?
The boundaries used have changed -
From Assisted to Sympathetic
AI: Is it an assistant or a replacement? What can and can't be done? These questions must be answered clearly.
In May this year, the "chest cleansing and multi-checking smart body" was implemented at Zhongshan Hospital, affiliated with Fudan University, and more than 2.5 million cases have been processed. Ye Xiaodan, chief physician of the radiology department at the hospital, said: "After multicenter validation, chest scanning and multi-spectral images reduced the screen time by 33 percent, and we really felt the burden reduction and efficiency benefits brought by the large model."
However, the potential for AI to cause misinformation or even misdiagnosis must also be taken seriously.
Professor Chen Sijiao from the First Affiliated Hospital of China Medical University cited a series of alarming cases: when questioning AI, Some patients judge chest pain to be bilateral nerve pain, some treat a childhood fever rash as an ordinary viral rash, and others misjudge skin lesions as benign moles when they are actually myocardial infarction, meningococcalemia, and early melanoma.
Zhang Bo, an academician of the Chinese Academy of Sciences and honorary dean of the Institute for Artificial Intelligence at Tsinghua University, offered a sober definition: AI lacks genuine reasoning and causal understanding capabilities; its "black box" nature and unexplainability are cause for concern, and it is not yet capable of assuming self-responsibility. He goes on to say that human doctors are inherently motivated, reason and understand, credible and accountable - "and that's why clinical advice given by large-language models ultimately needs to be handled by human doctors."
Zhang Bo has defined the role boundaries for physicians in the AI era: doctors are both clinical decision-makers and safety overseers. The relationship between AI and doctors is not one of replacement, but of symbiosis, because“The ultimate decision - "what to do" - still belongs to the person who has the warmth, experience and responsibility. In the foreseeable future, AI will not replace doctors; instead, it will demand higher standards from them. Proficiency in understanding, utilizing, and explaining AI will become fundamental competencies for physicians of the future.”。
Liu Zhenlong defines the boundaries of AI from another dimension:“Technologically, AI can perform exceptionally well, but doctors face individuals with thoughts and emotions. First, we must understand the patient's needs and expectations, engaging in eye contact and emotional exchange with them; AI is unlikely to fully replicate this aspect of care in the short term. The doctor's job is - sometimes, to heal; Often, help out; Always, comfort yourself - the machine needs to be improved.”He also pointed out that vigorously developing medical AI "is not a choice question, but a must-do question, and if you don't keep up with the times, you will definitely fall behind." (Looking ahead to a new era)