Medlinker's Decade-Long Commitment: Future Doctor Leverages AI to Solve the Scalability Challenge of High-Quality Healthcare

On September 16, Dr. Wang Shirui, Founder and CEO of Medlinker Group, provided an in-depth analysis of the development path, core technologies, and industry value of its AI product, “Future Doctor,” at a small-scale media briefing held in Beijing. With over a decade of deep engagement in the healthcare sector, Medlinker is leveraging MedGPT as its engine to utilize AI technology in breaking the “impossible triangle” of high-quality, affordable, and accessible medical care, thereby enabling the scalable provision of premium healthcare services.

Dr. Wang Shirui, Founder and CEO of Medlinker

According to Dr. Wang Shirui, Medlinker’s development trajectory has always centered on addressing core pain points in healthcare. Starting as an online physician community in 2014, it became one of the first enterprises in China to obtain an internet hospital license in 2017, and by 2020, it had grown into the largest online chronic disease management platform in China. As early as 2017–2020, Medlinker explored the application of AI in the medical field, but the project was forced to stall due to technological limitations at the time.

With the advent of ChatGPT, Medlinker recognized the application potential of general-purpose large language models in healthcare scenarios. In May 2023, Medlinker launched MedGPT, a medical large language model powered by proprietary algorithms. Within one month of its launch, it collaborated with West China Hospital to complete the world’s first double-blind clinical trial assessing the diagnostic consistency between an AI physician and human physicians. Among 91 valid cases, MedGPT demonstrated a 96% diagnostic consistency rate with ten attending physicians from top-tier (Grade A tertiary) hospitals. This trial was conducted a full year earlier than similar tests for ChatGPT-4.

In August 2025, the Future Doctor app, designed for consumer-end users, was officially launched. Concurrently, related research findings were published in an American academic journal, and the company became the only Chinese enterprise nominated in the Best Digital Health Solution category of the Galien Awards (hailed as the “Nobel Prize of the U.S. pharmaceutical industry”), garnering international recognition for its capabilities.

The low tolerance for error in healthcare dictates that AI technology must overcome the “hallucination” bottleneck, which constitutes the core competitiveness of MedGPT. According to Dr. Wang Shirui, MedGPT employs a proprietary “fast-slow dual-system” architecture: the fast system leverages multiple foundational large language models to achieve natural semantic interaction, addressing the challenge of “understanding human language”; the slow system, based on Retrieval-Augmented Generation (RAG) technology, validates and corrects the output of the fast system by drawing on proprietary medical knowledge graphs, expert consensus, and other resources, thereby thoroughly mitigating the risk of general-purpose AI “confidently generating nonsensical information.”

The training foundation of this architecture is immense: 8 million clinical diagnosis and treatment records, 2 billion doctor-patient dialogue logs, and 50,000 of the latest global medical guidelines and authoritative literature, covering more than 1,200 clinical pathways. This ultimately enables diagnostic and therapeutic capabilities across 26 specialties and over 30,000 diseases—a scope that essentially encompasses all common and complex conditions encountered in clinical practice.

In the 2025 upgrade tests, MedGPT comprehensively outperformed well-known general-purpose large language models, both domestic and international, such as Deepseek-R1 and Gemini-2.5 Pro, across 30 safety and efficacy metrics established based on China’s medical malpractice classification standards, thereby demonstrating its leadership in the vertical healthcare sector through data.

Dr. Wang Shirui stated that MedGPT’s technology selection stems from a profound understanding of the limitations of first-generation AI. Although early AI could perform logical reasoning, it failed to achieve natural language understanding, making it difficult to handle the ambiguous expressions common in real-world clinical practice. While the emergence of the Transformer architecture resolved issues related to language comprehension, it lacked the capability for logical verification. “The dual-system approach of ‘fast and slow’ thinking mimics the human cognitive pattern of ‘intuitive proposal + logical verification,’ enabling AI to both understand natural language and make precise decisions.” More critically, this training model, based on rule bases and clinical guidelines, avoids the high costs associated with large-scale data acquisition. By leveraging the technical expertise of its core R&D team, the system achieves traceability and controllability.

“No matter how advanced the model becomes, a physician’s clinical experience remains an irreplaceable core value,” emphasized Dr. Wang Shirui on site.

Therefore, future physicians will not seek to rely solely on AI for diagnosis and treatment, but rather create an “intelligent digital twin of the physician”—enabling AI to learn from experts’ clinical experience, decision-making frameworks, and even communication styles, thereby helping physicians become “super-individual practitioners in the era of intelligent healthcare.”

Currently, Future Doctor has collaborated with over 40 experts at the level of chairpersons and vice-chairpersons of the Chinese Medical Association to conduct joint research, covering nearly all core clinical specialties. These leading experts contribute daily case reviews and supplement guideline-excluded clinical experience, creating a “data flywheel” that continuously optimizes model performance.

Building on this foundation, Future Doctor has launched two AI agents: a consultation and diagnostic agent for consumers that provides clinical-grade medical decision-making by integrating past medical history; and a patient management agent for patients with cancer and autoimmune diseases, capable of real-time monitoring of drug side effects and dynamic adjustment of treatment plans. A patient with lung adenocarcinoma and bone metastases successfully achieved two core health goals after three months of intelligent management, demonstrating the practical value of such services.

To ensure medical quality, WeDoctor has established a three-tiered defense system: strictly adhering to evidence-based medicine to reject overtreatment; conducting multidisciplinary team (MDT) consultations for all cases, thereby overcoming the limitation of MDTs in physical hospitals that typically cover only complex cases; and having final diagnostic and treatment opinions reviewed and signed by over 6,000 physicians licensed to practice in internet hospitals, who assume legal liability.

This “AI-assisted + physician-endorsed” model elevates online serious healthcare to new heights.

"Leveraging high-frequency doctor-patient interactions, WeDoctor has designed a comprehensive business ecosystem covering both B-side and C-side users."

C-end users can access the full spectrum of services—“consultation, diagnosis, medication purchase, testing, and appointment registration”—via the app. By integrating with supply chains from platforms such as Meituan and JD.com, the service enables medication delivery to patients’ doors within 30 minutes. It currently addresses 95% of medical consultation needs, helping patients reduce treatment costs by nearly 90%.

The B2B segment serves entities such as hospitals, pharmaceutical companies, and insurance firms: it provides underwriting assessments for individuals with pre-existing conditions to Junlong Life Insurance; embeds consultation benefits into Ping An Insurance’s offerings; partners with OPPO to integrate health services; and delivers pre-consultation triage, intra-consultation diagnostic support, and post-consultation follow-up services to hospitals. In 2024, B2B revenue accounted for two-thirds of the platform’s total revenue.

Regarding the profitability model, Dr. Wang Shirui stated that funding for AI research and development currently comes primarily from profits generated by mature business lines, such as the company’s internet hospital. This has provided the patience needed for technological iteration without “forcing growth.” In the long run, as physicians become the primary distribution channel for healthcare service traffic, commercial value will naturally emerge. However, the core objective remains “promoting the accessibility of healthcare, ensuring that the general public can access, afford, and conveniently receive medical care.”

Regarding industry challenges, Dr. Wang Shirui, who has been deeply engaged in the medical sector for many years, maintains a clear-headed perspective: data silos are not an insurmountable obstacle, as MedGPT is trained based on rules rather than massive volumes of unstructured data; while technical hurdles remain in multimodal recognition, diagnostic accuracy can still be ensured through physician review. The future development path for physicians is now clearly visible: evolving from “replicating doctors” to “replicating medical service systems,” integrating supply chain resources through traffic entry points to simulate a complete medical ecosystem.

“We aim to build an online hospital free from independent KPIs, where all recommendations are based on clinical guidelines and experience rather than vested interests.” Dr. Wang Shirui’s words reflect Medlinker’s long-standing commitment to the essence of healthcare. Empowered by AI technology, Medlinker is striving to break down geographical and hierarchical barriers to high-quality medical resources, ultimately ensuring that “patients no longer face difficult choices.” This is not only a commercial blueprint for future physicians but also a direction for the healthcare industry toward greater equity and efficiency.