446 million people open AI apps every month to ask questions — this is no longer an "early adopter" crowd, but mainstream Chinese internet users. Understanding who they are, how they ask, and which sources they trust is the prerequisite for any AI search optimization (GEO) strategy.
QuestMobile's Q1 2026 AI Application Insights (published 2026-04-21) provides the full picture: China's AI native apps reached 446 million MAU in March 2026, adding over 130 million users in a single quarter, with 173.3 minutes of average monthly usage per user, up 30.3% since November 2025. This article breaks down the data to answer one question marketers care most about: what content do AI users actually want?
1. Who Asks, and Where: The Composition of 446M MAU
Headline Platform Landscape (March 2026)
| Platform | MAU | Company | Core user profile | Source preference |
|---|---|---|---|---|
| Doubao | 345M | ByteDance | Mass-market, deep lower-tier penetration | Accessible, scenario-based content |
| Tongyi Qwen | 166M | Alibaba | Male-skewed | Technical specs, industry data |
| DeepSeek | 127M | DeepSeek | Tech-savvy, efficiency-driven | Rigorous, traceable content |
| Kimi/Yuanbao/Ernie | — | Moonshot/Tencent/Baidu | Yuanbao strong in developed cities | WeChat ecosystem content |
Behavioral Data: Habits Are Set
- Doubao: 54.8 uses per person per month, up 22 year-over-year — near-daily usage
- DeepSeek: 41.7 uses per person per month
- The 2026 Spring Festival drove 130 million new AI users; post-holiday retention validated platform value
Source: QuestMobile Research Institute, Q1 2026 AI Application Insights (published 2026-04-21).
2. How User Profiles Shape AI Source "Taste"
QuestMobile explicitly notes: different platforms' user profiles determine their content source preferences. This is a crucial operational signal — the same content has very different citation probabilities across platforms.
Qwen: Tech-Skewed Users → Hard Content Wins
Qwen's users are male-skewed and ask technical, parameter-driven questions. To optimize for Qwen, publish product specs, technical documentation, and white-paper-style content so your brand becomes the source for "parameter-type" answers.
Yuanbao: Developed-City Users → WeChat Ecosystem Is the Battlefield
Yuanbao users concentrate in developed cities, and its source preferences tie deeply into WeChat official account content. Brands should build a WeChat content matrix — optimized headlines, structure, and opinion density — so WeChat articles become citable sources.
Doubao: Mass-Market Users → Accessible, Scenario-Based Content
Doubao's 345M MAU spans the broadest audience, asking "how to choose, how to use, how much it costs". Content for Doubao must be conversational, scenario-based, and conclusion-first — no jargon stacks.
3. Four Takeaways for GEO from User Insights
Takeaway 1: Customize Content per Platform, Don't One-Size-Fits-All
Profile differences mean a single content strategy wastes half the opportunity. Build a "one source, multiple forms" mechanism: technical version (Qwen), WeChat version (Yuanbao), and accessible version (Doubao) for the same topic. See our GEO optimization guide for deployment methods.
Takeaway 2: Decision Questions Are the Highest-Value Scenario
Behind 173.3 minutes of monthly usage lies a flood of "who should I choose" questions. Comparison and checklist content built around decision scenarios is the fastest route to higher recommendation probability.
Takeaway 3: Spring Festival Growth Validates the "Node Dividend"
The 130 million new users during Spring Festival show AI adoption is still accelerating, and new users flush AI answers with brand-new question pools. Continuous content output is required to keep pace with answer iteration.
Takeaway 4: In the Retention Era, Source Trust Outweighs Traffic
Retention validates platform value; for brands it means being cited once is not enough — be cited consistently and positively. Source authority, attribution, and data backing (the EEAT principle) determine recommendation quality. See our analysis of LLM citation mechanisms.
4. FAQ
Q1: Do SMBs need to optimize for every platform separately?
Prioritize: cover Doubao first (largest user base) and Qwen (decision-oriented users), then decide on Yuanbao (WeChat ecosystem) based on your industry. Use "one source, multiple forms" to control costs when budgets are tight.
Q2: How do we assess our current AI visibility?
Ask brand and industry keywords in Doubao, Qwen, and DeepSeek; record whether your brand is mentioned and whether the context is positive. Do this monthly at a fixed time to build a baseline.
Q3: Will user profiles stay the same?
No — they evolve. QuestMobile's quarterly reports show dramatic shifts (Qwen rose from TOP6 to TOP2 in one quarter). Keep tracking user and platform dynamics, and review strategy every six months.
Q4: Can we cite these figures in our own content?
Yes, but verify through QuestMobile's official channels and attribute the source. Data-backed content is itself preferred by AI source selection — a good way to raise LLM inclusion probability.
Written by Zheming Digital Communication Research Institute. Data cited from QuestMobile Research Institute public reports (published 2026-04-21). AI search optimization and GEO consultation: +86 18917757529 ｜ jaysun@widesight.cn.