China's AI App Semi-Annual Report: Douyin's Doubao Leads with 382 Million MAUs, Alibaba's Qianwen Surges 58-Fold

China's AI application market has seen over a year of intense competition, and the landscape of leading players is now taking shape. According to the latest "2026 First-Half AI Application Market Development Insight Report" released by third-party data agency QuestMobile on July 14, the MAUs of China's AI-native apps approached the 500 million mark as of May 2026. ByteDance's "Doubao" firmly leads with 382 million MAUs, while Alibaba's "Qianwen" and the startup-developed "DeepSeek" rank second and third, respectively, forming a first tier of apps with over 100 million MAUs.

The report indicates that over the past one to two years, major large language model (LLM) companies engaged in a high-intensity "user acquisition war," capturing users' screens through massive advertising campaigns and pre-installation partnerships with smartphone manufacturers. The dust is now settling on this traffic battle. The growth model of relying solely on spending to attract new users is changing, and users are rapidly consolidating around a few leading applications.

Data shows that as of May 2026, Doubao had 382 million MAUs, maintaining a clear lead in absolute terms. Qianwen ranked second with 167 million MAUs, and its year-over-year growth rate reached a staggering 5,792.9%, demonstrating a remarkable ability to expand its user base. DeepSeek ranked third with 130 million MAUs, securing its position primarily through technical reputation and organic growth within the developer community.

It is noteworthy that the usage scenarios for AI applications are undergoing a fundamental shift. The report's statistics on user scale across five types of AI application channels reflect a substantial change in user interaction habits. AI-native apps now have 499 million MAUs, a year-over-year increase of 85.4%, making them the fastest-growing channel. Built-in AI assistants from device manufacturers like smartphone makers have the broadest reach, with 755 million MAUs, a 14.0% year-over-year increase. AI application plugins have 644 million MAUs, a 7.9% year-over-year decrease.

On the PC side, web-based AI applications have 172 million MAUs, a significant year-over-year decline of 22.8%. Meanwhile, PC client-based AI applications, though starting from a smaller base, have only 18 million MAUs but saw a 20.1% year-over-year increase. The contrasting rise and fall in this data clearly reveal a trend: early AI tool experiences were mostly dependent on PC web browsers or existed as third-party software plugins. Now, standalone AI-native apps and system-level AI features integrated by smartphone manufacturers have become mainstream. The decline in PC web activity confirms that AI tools are shifting from low-frequency "trial" scenarios to daily, high-frequency use centered on mobile devices.

After the initial phase of capturing user scale, engagement and depth of use have become core metrics for measuring business health. The report's data shows that as of May, the five types of AI applications exhibit significant differences in user stickiness. AI-native apps lead by a wide margin with 92.7 average monthly uses per person. AI application plugins follow at 60.9 uses, device manufacturer AI applications at 51.4 uses, and PC clients and PC web versions trail significantly at 26.9 and 25.3 uses, respectively.

Below is a comparison of key data for the five AI application channels:

Application Channel Type	MAU Scale (100 Million)	Year-over-Year Growth	Avg. Monthly Uses Per Person
Device Manufacturer AI Apps	7.55	+14.0%	51.4
AI Application Plugins	6.44	-7.9%	60.9
AI-Native Apps	4.99	+85.4%	92.7
PC Web Versions	1.72	-22.8%	25.3
PC Clients	0.18	+20.1%	26.9

Note: Data as of May 2026, sourced from QuestMobile's "2026 First-Half AI Application Market Development Insight Report."

This comparative data reveals a key insight: although built-in AI applications from device manufacturers have achieved a massive 755 million MAU scale thanks to pre-installation and underlying system permissions, their average monthly usage frequency per person is significantly lower than that of AI-native apps, which require users to actively download them. This implies that while built-in AI significantly lowers the barrier for users to access LLMs, it has yet to demonstrate stronger stickiness than independent AI-native apps in terms of tapping into real, essential-use scenarios and deep user retention.

Looking at the competitive dynamics among leading apps, Doubao's leading position is attributable to its early start and sustained, large-scale marketing investment. ByteDance has clearly tilted significant resources toward the AI field over the past year, directing traffic to Doubao through its super-app matrix, including Douyin and Toutiao, and conducting intensive advertising across multiple channels.

Qianwen's explosive growth reflects Alibaba's aggressive pursuit in the AI race. Its 58-fold year-over-year growth rate illustrates a trajectory of rapid expansion from a very low base. The support of Alibaba Cloud as underlying infrastructure, along with integration into ecosystems like Taobao and DingTalk, has provided Qianwen with a broad space for real-world application.

DeepSeek, holding steady in third place with 130 million MAUs, has followed a different growth path from the top two. This relatively young company relies more on its technical prowess and open-source strategy, accumulating significant goodwill within the developer community and achieving relatively low-cost user growth through community-driven dissemination.

The report points out that with the MAU scale of AI-native apps approaching 500 million, China's AI application market has moved past its pioneering phase of attracting traffic through hype. Leading players have not only widened the MAU gap with mid-tier competitors but have also clearly shifted their competitive focus to depth metrics such as per capita usage frequency and user retention.

As infrastructure and foundational LLM capabilities gradually converge, market performance in the next phase will directly depend on each company's ability to analyze specific business scenarios and execute engineering solutions. The era of simply competing on model parameter scale or marketing budgets is passing. The key to victory will be establishing user habits in high-frequency, essential-use scenarios.

Industry insiders analyze that while the first-tier landscape of China's AI application market is initially stable, Qianwen's high-speed growth and DeepSeek's technical reputation mean the competition is far from over. Furthermore, although device manufacturer AI applications lag in user stickiness, their massive base of 755 million users and system-level integration advantages make them a potential variable that cannot be ignored. In the coming year, each player's performance in deep scenario cultivation and commercial monetization will be the core focus for observing the market's direction.