In Depth: China’s Tech Giants Burn Cash to Try to Dominate AI Health Care

1. In late 2025, Ant Group began an aggressive advertising campaign for Afu, its rebranded personal health AI assistant, spending hundreds of millions of yuan within a month. By early 2026, OpenAI launched its own version, ChatGPT Health, inviting select users to a beta test. The proliferation of these AI health assistants is mirrored by a surge in competing Chinese products, such as Future Doctor (backed by Medlinker), JD Health’s Dawei, Alibaba’s Quark, Baidu’s Ernie Health Manager, Iflytek’s iFlyHealth, and ByteDance’s Xiaohe AI Doctor. These initiatives reflect a strategic focus by China’s internet giants, health care service companies, and emerging AI startups to capture a share of the expanding personal health market.[para. 1][para. 2][para. 3]

2. China’s state regulators have responded to this boom. In October 2025, the National Health Commission, in conjunction with four other departments, issued guidelines for promoting and regulating AI in healthcare, targeting the foundation of several pilot national AI application bases in medicine by 2027. Major cities and provinces such as Beijing, Shanghai, Zhejiang, Henan, and Hefei have subsequently outlined their own plans for pilot projects, underlining the sector's growing strategic importance and regulatory supervision.[para. 4]

3. AI’s prowess in medicine has been demonstrated through competitions like the one held at Beijing Arion Cancer Center on December 20, 2024. There, a multidisciplinary team of veteran doctors competed against five AI models: Doubao, Baichuan, Xiaohe, ChatGPT o1, and Gemini. Judges scored each participant over six clinical rounds, including diagnostic and treatment planning skills. ChatGPT o1 outperformed all, scoring 314.5 points and taking first place in five rounds. Gemini secured second with 265 points, while the human doctors followed with 251.5 points. Xu Zhonghuang, one of the participants, noted ChatGPT o1’s balanced capabilities, efficiency, and impressive provision of holistic care, including nutritional and psychological support, though human doctors excelled in more nuanced treatment planning. Xu expressed anxiety over the rapid progress in AI, predicting unpredictable advancements over the coming years.[para. 5][para. 6][para. 7][para. 8][para. 9][para. 10][para. 11]

4. The competitive edge of AI was further validated as AI models began to reliably pass medical licensing exams. OpenEvidence’s AI achieved a perfect score on the U.S. Medical Licensing Examination (USMLE) in August 2025, while Chinese companies like iFlytek, Quark Health, and Baichuan Intelligence announced their models had passed China’s National Medical Licensing Examination. Such feats have unsettled the medical community; hospital directors who lost to AI in competitions were uncomfortable publicizing their defeats, and junior doctors expressed anxiety about being replaced.[para. 12][para. 13][para. 14]

5. For consumers, consulting AI has rapidly become routine. People increasingly use large language models—such as DeepSeek, ChatGPT, Doubao, or Kimi—for health-related issues, benefiting from rapid and detailed responses on symptoms, suggested diagnostics, treatments, and even follow-up care. The preference for AI over traditional hospital visits or unreliable online searches is driven by the convenience, perceived objectivity, and clarity provided by AI interfaces, despite prominent disclaimers that content is for reference only.[para. 15][para. 16][para. 17][para. 18]

6. Recent usage statistics underscore this trend: As of January, OpenAI reported that over 5% of ChatGPT’s global queries relate to health care, totaling billions per week. Among more than 800 million active users, one in four asks a medical question each week, and over 40 million people make medical inquiries daily. Frequency is higher in areas with limited medical access, with rural users sending an average of 600,000 health-related messages weekly—70% occurring outside normal clinical hours.[para. 19][para. 20]

7. Specialized consumer medical AI products claim advantages over general-purpose models, including medical-specific training data, expert-like reasoning (using multi-round questioning), reduced hallucinations, and improved accuracy and efficiency. These products aim to provide direct diagnostic and treatment advice, differentiating themselves in a rapidly crowding field.[para. 21]

8. Despite significant investment in technology, data, and expert collaborations, the industry faces substantial uncertainty regarding profitability. Current consumer-facing AI health consultations are mostly free, and business models remain elusive. Experts suggest that as general large models continue improving, performance gaps will narrow, calling into question the viability of specialized vertical products. Chen Liang, Ant Group’s senior vice president, admitted the intense internal debate over business models but maintained that as the population ages and AI products create societal value, a viable business path will eventually emerge.[para. 22][para. 23][para. 24][para. 25][para. 26]