Baidu has worked out what to put underneath an AI health answer: a real doctor. At an event in Beijing on 24 October, Baidu Health upgraded its Wenxin Health Manager (文心健康管家) and launched a “Good Doctor Matrix” (好医矩阵) and a “Health Camera” (健康相机). Search is at the centre of it. Users who search health questions in the Baidu App now get a dual layout that Baidu itself describes as “AI answer above, doctor entry below” (上面AI回答、下面医生入口).
Baidu says more than 130 million users search health questions on Baidu every day, and its AI health services now have more than 50 million monthly active users. Those two numbers explain why this matters for anyone in health publishing or healthcare marketing in China.
What Baidu launched
Based on reports from China National Radio (CNR) and Sina Finance:
- Good Doctor Answers (好医答): a consultation room with more than 150,000 doctors from tertiary (三甲) hospitals. It runs 24/7, with a fastest response time of one minute, and is reached directly from Baidu App search for “百度健康” or a health question. The AI first organises the user’s symptoms, test results, duration and history, then routes them to a doctor in the right department, who asks follow-up questions and gives the actual advice. The search-entry version currently covers four departments: dermatology, internal medicine, traditional Chinese medicine and gynaecology. More will open department by department.
- Good Doctor Extra Slots (好医加号): a booking bridge to offline care. Patients upload records and preferred dates, and the doctor personally reviews and grants extra appointment slots. Baidu claims coverage of more than 80% of public tertiary hospitals, 50,000+ participating doctors, 120,000+ extra slots a day, a 70% approval rate, and nearly 6 million successful bookings so far.
- Health Camera (健康相机): multimodal input. As well as typing, dialect voice and photos, users can now shoot video and talk while filming, for things like skin conditions or movement problems. Baidu says the model was trained on 3 million professional multimodal images and more than 100 million diagnostic-logic records. Its report-reading feature has been used 27 million times in the past year, with claimed accuracy above 98%.
- Open Plan (好医开放计划): chronic-disease apps, health hardware, insurers, pharmacy O2O and employee health services can plug into Baidu’s doctor and AI resources via MCP, API or H5. A separate Good Doctor Alliance (好医联盟) targets doctors, medical societies and hospitals.
The user-trust number that shaped the product
The most useful line in CNR’s report is Baidu’s own research finding: nearly 70% of AI health users don’t fully trust AI’s direct answers and cross-check them, and when a child or a serious illness is involved, they still want a doctor. Baidu Health’s general manager Yang Minglu put the philosophy as AI that “empowers but does not replace, assists but does not lead”, with the decision always “in the hands of professional doctors”.
That is a big shift from the Baidu of the 2016 Wei Zexi scandal, when paid medical listings were at the heart of China’s biggest search-trust crisis. Ten years later Baidu is building accountability into the SERP layout itself.
Why this matters for search
Look at the query flow. A user asks Baidu a symptom question. They get an AI summary, and directly under it an entry point to a verified tertiary-hospital doctor who can answer in a minute and potentially book them an appointment. Where does the third-party health site fit in? On the highest-intent medical queries it is now competing with a Baidu-owned answer and a Baidu-owned human.
We have seen this pattern before. Baidu built Zhidao, Baike and Baijiahao to keep answers in its own ecosystem. Baidu Health is the same idea for the most regulated vertical there is, now with an AI layer on top.
For hospitals and clinics, the opportunity runs the other way. Doctor profiles, the 好医榜 and 好医评 doctor ranking and review features and extra-slot availability become the local-search inventory that matters. The MCP/API Open Plan also suggests Baidu wants its health stack to be the default tool that other agents call, not just a destination.
Dan’s take
I’ve long argued that YMYL is where AI answers either earn trust or lose it, and Baidu has made the honest call: in health, an unaccountable AI summary isn’t good enough. Putting a named, credentialed person under the answer is a better design than anything Google offers in AI Overviews today. It is also easier for Baidu, because it has spent a decade building a doctor network that Google simply doesn’t have.
The evidence caveat: almost every number here, from 50 million MAU to 98% accuracy and 6 million bookings, is Baidu’s own claim from a launch event. There is no independent verification, and “accuracy” in report reading has no published methodology. I’d treat the direction as solid and the figures as marketing until we see third-party data.
The practical takeaway for health brands working in China: assume clicks on informational symptom queries keep falling. Put effort into what Baidu routes to: verified doctor presence, hospital listings and structured, authoritative content that the AI layer can cite. Expect more of this at Baidu World on 12 November in Beijing. More in Baidu AI.