Vertical AI Apps Like Ant Afu and Doubao Aixue Now Rival Yuanbao: The 2026 Brand Visibility Blind Spot

Two vertical AI apps — Ant Group's health assistant Ant Afu (28.97 million monthly active users) and ByteDance's study tutor Doubao Aixue (19.44 million) — together reach 48.41 million people a month. That is 97% of Tencent Yuanbao's 49.84 million, a general-purpose assistant most international brands already monitor. Almost no international brand tracks either vertical app. Yet in health and education, these apps, not Doubao or DeepSeek, are where the buying question gets answered.

This analysis uses QuestMobile's H1 2026 AI native app rankings (June 2026 data, reported by HK01 on August 6, 2026) and Ant Group's WAIC 2026 disclosures. Brand-score commentary refers to the category patterns tracked on the hubGEO /brands page.

The Ranking Most Brand Teams Stop Reading at Rank Three

QuestMobile's H1 2026 report, published July 14, puts Doubao, Qwen and DeepSeek in a clear "first tier." The June 2026 MAU figures look like this:

(Some syndicated write-ups of the same report print Doubao's figure as "3.82 billion"; the 382 million figure matches the app-level ranking and is the one used here.)

Two things stand out. First, Doubao's user base is roughly 2.3 times Qwen's, which is why most GEO budgets default to Doubao. Second, ranks five and six are not general assistants at all. They are single-purpose apps with an audience larger than Kimi's by a factor of four (Afu) and nearly three (Aixue). Kimi, which many Western brand guides still treat as a core China AI channel, sits below both.

The overall market context matters too. QuestMobile counts about 499 million monthly users across all AI native apps (up 85.4% year over year), each averaging 92.7 sessions and 183 minutes a month. The market is large enough that verticals can build real scale without being "niche."

Why Ant Afu Is a Different Kind of Answer Engine

Ant Afu is not a chatbot that happens to answer health questions. At WAIC in July 2026, Ant Group said Afu has passed 100 million users, handles more than 10 million health consultations a day, and connects to over 5,000 hospitals and 300,000 real doctors. It covers Q&A, report interpretation, online consultation, appointment booking and medical-insurance payment in one flow.

Its distinguishing feature is "AI Q&A plus doctor review": answers can be checked by doctors, and more than 2,000 doctors have built AI avatars led by eight academicians. For a brand, this changes what "getting cited" means:

• The citation filter is clinical, not editorial. A general assistant like Doubao will happily cite a Xiaohongshu review. A health app that advertises physician verification has strong incentives to prefer hospital data, clinical guidance and regulated sources.
• The funnel is closed. Because Afu can book appointments and process insurance payments, a recommended product, clinic or service can be one tap from purchase. Our earlier analysis of Doubao's hotel commissions showed that closed-loop recommendation tends to concentrate choices into a very short list.
• The audience is already in a health decision. Ten million daily consultations is a high-intent query stream, closer to a pharmacy counter than a search box.

The 100 million cumulative users against 28.97 million monthly actives suggests that roughly 29% of the user base is active in a given month. That ratio is our own back-of-envelope estimate, since the two figures come from different sources and definitions, but it indicates the app has real retention rather than a one-off install spike.

Doubao Aixue and the Parent-as-Buyer Problem

Doubao Aixue, at 19.44 million MAU, is the sixth-largest AI native app in the country. It sits inside the Doubao family, so it inherits ByteDance's distribution, but its queries are of a different type: homework help, subject explanations, study-plan questions. The commercial surface is indirect. Students ask, but parents pay for tablets, learning devices, tutoring services, nutrition and even eyesight-protection products.

Education-adjacent international brands, from learning-device makers to children's nutrition and stationery, typically measure themselves in Doubao's general assistant. That is the wrong test. A brand can score reasonably in the general model while being absent from the study-tutor context where the product recommendation is more natural.

Three Implications for Visibility Measurement

1. Your brand score is a property of a surface, not a model family. Doubao, Doubao Aixue and Doubao inside Douyin share a name and a parent company, but they draw on different query sets and different retrieval rules. The same logic that explains why DeepSeek and Qwen cite different sources (covered in our source attribution divergence analysis) applies inside a single company's product line.

2. Vertical apps reward verifiable proof, not volume. In a health app with physician review, ten additional blog mentions do less than one entry in a hospital system, a registered clinical study or a regulator's database. Brands in supplements, skincare, infant products and medical devices should expect the weight of source authority to rise sharply. Our work on low-authority source ratio penalties is a useful starting reference.

3. Compliance is part of visibility. Health claims in China are tightly regulated, and AI answers that repeat unapproved claims create risk for the brand. See our overview of GEO and advertising-law liability before building any health-category content program.

A Quick Exposure Check by Category

How This Fits the Wider Concentration Story

The top-three concentration is real: Doubao, Qwen and DeepSeek hold the first tier by a wide margin, and a 382 million versus 49.84 million gap to fourth place shows how steep the drop is. But steep concentration at the top does not mean uniform influence. A 10 million-user vertical app with a purchase flow attached can move more revenue in its category than a 100 million-user general assistant where the same question is one of thousands. Engagement points the same way: QuestMobile reports that 30.0% of DeepSeek's users and 27.5% of Doubao's spend more than ten minutes per session, which tells us users are doing real research in these tools, not just quick lookups. Vertical apps are likely to be at least as deep, because a health consultation or a study session is inherently longer than a factual query.

For agencies, this also changes reporting. A client dashboard showing one blended "China AI score" hides the surface where the category is actually decided. Ask vendors to break results out by app and by query intent, and treat any vendor that cannot do so as measuring the average rather than the answer your customer sees.

What We Cannot Yet Say

Public data on these two apps is thin. QuestMobile discloses MAU, not citation sources or brand-level answer share, and Ant Group has not published how Afu selects products or sources. We have not yet run a systematic brand-score panel against the vertical apps, so any claim that a specific brand scores higher or lower there would be speculation. What the numbers do justify is treating the surfaces as measurable and testing them directly.

Takeaway for Brand Marketers

Stop treating "China AI visibility" as one number from three general assistants. If you sell in health, wellness, infant care or education, do this in the next 30 days:

1. Audit the vertical surfaces. Run 30–50 category and symptom or study-intent queries in Ant Afu and Doubao Aixue, and record whether your brand appears, how it is described, and which sources are cited.
2. Fix authoritative records first. Make sure product registrations, clinical references, hospital or professional endorsements and Baike entries are accurate and consistent. These are the sources a physician-reviewed app can rely on.
3. Reallocate part of the Doubao budget. If your category matches a vertical app, test spending on authority content for that surface before adding more volume in the general assistant.
4. Brief legal and the agency together. Any health or child-related claim needs sign-off before it is published for AI retrieval.

The headline MAU table rewards the biggest apps. The purchase decision, in health and education, is increasingly made in the smaller ones.

Sources: QuestMobile H1 2026 AI app report via IT之家; HK01, AI native app rankings (Aug 6, 2026); IT之家 on Ant Afu at WAIC 2026.

Related: Brand scores by model on /brands · Financial services brands in China AI · Doubao hotel commissions and recommendation attribution