China’s Medical AI Battle: A Decade of B2B Pain—Now Turning Toward Patients

On February 16, “Ant Afu” (蚂蚁阿福) appeared on China’s Spring Festival Gala in a way that looked almost un-tech: not a futuristic humanoid robot, not a general-purpose AI demo, but a familiar, everyday scenario—elderly checkups and chronic disease management.

That placement wasn’t accidental. It was one scene in a much larger strategy: occupying user mindshare at scale.

From December to now, Afu has shown what happens when a top-tier operator decides to push into an industry with full force. In December alone, Ant Group reportedly spent “several hundred million” RMB on marketing. The result was a highly coordinated campaign across:

• “Serious” channels (KOL recommendations and professional content)

• Lower-tier cities and mass-market promotion (cash incentives for app downloads)

• Broad exposure via direct and indirect nationwide channels, including the Spring Festival Gala

This all-out online and offline push widened the very top of the user growth funnel. By the 23rd, Afu’s app reportedly surpassed 100 million users, with 52% of new users during the Spring Festival coming from third-tier cities and below.

In many ways, “going fully to C” (direct-to-consumer) is the signature story of this round of collision between AI and healthcare.

Ant Afu and Baichuan Intelligent—after weighing options—both chose to face users directly. And Afu’s high-profile “raise the flag and charge” approach reminded hospital leaders of something they know well: how Alipay once broke into the seemingly unshakable banking landscape and created mass adoption in payments.

Ant Afu

The first decade: Medical AI tried to change doctors (B2B)

For roughly the last ten years, most medical AI in China was a B2B game:

• Finding channels to enter hospitals

• Bundling with large equipment

• Partnering with local governments for population screening

• Sharing patient acquisition revenue with labs, imaging centers, private hospitals, and other commercial providers

These were expensive, hard-to-copy “one-off deals” that rarely became scalable moats.

Across waves—health education, appointment booking, online consultations, imaging AI—healthcare repeatedly appeared as a promising landing zone for technology hype. But after more than a decade of push and pull, a conclusion hardened:

China’s hospital-centered healthcare system—like a fortress—has limited interest in information technology that talks about “disruption,” even if it has money.

Healthcare is both:

1. a public livelihood issue that gives people a strong “doing something meaningful” narrative, and

2. an extremely difficult business where “doing the right thing” does not guarantee commercial returns.

Many participants entered with ambition and left disillusioned.

Now the question is shifting.

The new decade: Medical AI tries to change patient pathways (B2C)

Last decade’s ambition was to change clinical habits. This time, AI is trying to change how people decide where to go, who to see, and what to do first.

Will this “ideal story” finally break through?

01. Creating (or manufacturing) demand

After three years of rapid progress, the large-model-driven wave of medical AI has entered formal policy vision.

Two key signals:

1. National Health Commission (Oct 2025)

   The Commission issued guidance to promote and regulate “AI + healthcare,” listing 8 directions and 24 key applications, and setting targets such as:

   • By 2027: build high-quality datasets and trusted data spaces for the health sector

   • By 2030: achieve near-full coverage of intelligent assistance tools in primary care

2. National Healthcare Security Administration (Jan 2026)

   The NHSA announced pilots for “Personal Medical Insurance Cloud,” aiming to build a “full-population, full-lifecycle, full-scenario” smart医保 management model, with evaluation expected by March 2027.

Together, these policies give legitimacy to both B2B and B2C deployment:

• Hospitals gain new KPIs for digitization and intelligence upgrades

• Within 1–2 years, 1.3 billion insured individuals may have foundational data assets (tests, diagnosis, treatment, health consumption) under a standardized framework—an infrastructure no single company could build through “influence” alone

If privacy-preserving use and authorization can be implemented (a technical and governance challenge, but not an impossible one), then a plausible future emerges:

AI + B-side clinical data + C-side behavior data could unlock connections to commercial insurance, wearables, personal health management, and other monetizable services.

Yet despite many companies sensing the wind direction, only Ant appears willing to invest heavily “regardless of cost.”

One insider described Ant’s speed: even while policies were still being drafted, Ant was already recruiting aggressively. Ant and Baichuan are seen as the two major players gambling on To C; their talent needs overlap. In extreme cases, entire rows of Baichuan staff reportedly received headhunter calls.

In November 2025, Ant Group CEO Han Xinyi announced the upgrade of the “Digital Healthcare Division” into a “Health Business Group,” making healthcare one of Ant’s three pillars alongside lifestyle and financial services.

A key point: health management does not require the strongest foundation model. Afu’s advantage is integration—using large models to unify Ant’s existing online medical services:

• Users ask health questions and receive answers

• Users get guidance on which hospitals, departments, and doctors to visit

• The downstream journey (appointments, payments, insurance, in-hospital channels) is where commercial models can appear

The real question is not whether the model is powerful, but whether Chinese consumers truly want to pay for health management today—and if not, whether demand can be created.

Afu’s mass advertising in December signaled seriousness. An imaging AI veteran noted: nobody can guarantee medical AI will work commercially, but it’s also too early to dismiss it. “The operators behind Afu are among the few in China who can explain a To C story clearly in a complex scenario.”

Han defined Afu’s success in consumer terms:

• When half of Chinese users think of Afu first when they have health questions

• When more than half of Afu users would recommend it to family and friends

By January, some hospitals reportedly began to worry this could become a “patients surrounding hospitals” strategy—where AI influences hospital selection and patient flow.

The industry agrees AI won’t replace doctors. But what AI might do to everything else inside hospital walls is still unclear.

02. Nobody paid: the hard truth of B2B medical AI

Few people remember that Tencent was once positioned as part of China’s “national AI team” for healthcare.

In 2017, after AlphaGo’s rise, China named four major AI platforms:

• Baidu: autonomous driving

• Alibaba: smart cities

• iFlytek: intelligent speech

• Tencent: medical imaging AI

Tencent’s “Miying” (觅影) became a high-profile attempt to use hard tech to reshape care. At the time, optimistic narratives borrowed the “man vs machine” template: if AI could beat the world’s best Go player, why wouldn’t it beat doctors?

Capital followed. In 2018, medical AI financing doubled to RMB 7.6B, across 91 deals. Many imaging and specialty-AI companies expanded rapidly. By October 2025, China’s regulator had approved 110+ deep-learning-based Class III medical device software products.

But the core business dilemma never went away:

Even if accuracy exceeds 90%, risk still sits with humans. Vendors feared that less trained doctors might copy AI outputs blindly, and any mistake becomes a serious liability and reputational event.

Hospitals also pose structural barriers:

• Clinical products require evidence and validation

• Sales often requires relationships and long procurement cycles

• Public hospital tendering processes are slow and formalized

And most importantly:

A useful tool for doctors is not automatically a product that hospitals can—or are allowed to—pay for.

One regional agent promoting imaging AI said the practical policy stance is: hospitals are encouraged to use AI to improve efficiency, but not allowed to pay extra.

In a highly homogeneous market, competitors can copy features in months. To enter hospitals, vendors often give products away for free. After years of price wars, imaging AI struggled to build durable moats or repeatable business models.

Then macro conditions tightened:

• Anti-corruption pressure increased

• Fiscal inputs declined

• Public hospitals faced greater budget constraints

• Hospitals preferred large listed vendors (less risk of bankruptcy)

Many veterans now caution against repeating the “2018 illusion”: high concept heat, heavy marketing and promotion spending, but weak long-term value.

In recent years, Tencent reduced its direct healthcare ambitions and focused more quietly on To B—cloud and algorithms as a vendor, not a standalone consumer health app.

If you would like support on healthcare strategy, partnerships, or market entry across the Greater Bay Area, GBAHA can help—please contact us at service@greaterbayhealthcare.com.