According to the latest survey by Citi Research, China's AI applications are currently in a critical transition period characterized by "widespread adoption but shallow usage." However, behavioral patterns among heavy users are already signaling a clear acceleration—the inflection point for AI's deep integration into economic life is approaching. The tangible impacts on efficiency improvement, labor markets, and corporate profits are expected to accelerate.
Citi Research's Innovation Lab conducted a specialized survey in March 2026 involving 1,800 respondents. The results show that 70% of respondents use AI chatbots daily, with an average usage time of 49 minutes. However, only 5% are heavy users who engage with AI for over two hours per day. Preliminary calculations by Citi indicate that AI has brought about an approximate 4.3% efficiency improvement for respondents, which translates to about 1.8% for the entire economy when scaled up.
The survey also reveals that China's AI chatbot market is highly concentrated among leading internet platforms. Doubao, under ByteDance, leads with a 79% user penetration rate and a Net Promoter Score (NPS) of 54.4, significantly outperforming competitors such as Tencent Yuanbao (42%/28.3) and DeepSeek (37%/36.1). Citi points out that the diverse applications of AI will drive users to utilize multiple products across different scenarios, reducing the risk of a single platform dominating the market. The competitive landscape is expected to remain fragmented in the long term.
On the policy front, the Ministry of Industry and Information Technology issued the "Interim Measures for the Management of Anthropomorphic AI Interaction Services" on April 10, emphasizing a "human-centric" principle and clarifying measures to prevent algorithmic exploitation. The AI governance framework is taking shape rapidly. Citi anticipates that deeper fiscal and tax reforms, such as consumption tax adjustments, may see substantial progress within the year, gradually establishing mechanisms for distributing AI-driven benefits.
**High Adoption but Shallow Usage Remains Prominent** A February 2026 report from the China Internet Network Information Center (CNNIC) indicates that as of the end of 2025, China's internet user base reached 1.13 billion, with an internet penetration rate of 80%. The number of generative AI users grew by 142% year-over-year, reaching 602 million. However, the average monthly usage time was only about 183 minutes (approximately 6 minutes per day), representing a very small proportion of the average daily online time of 4.6 hours.
Revenue data further supports this conclusion. According to data from the Ministry of Industry and Information Technology cited by Xinhua News Agency, the core AI industry revenue in 2025 was approximately 1.2 trillion yuan, accounting for just 0.9% of the annual GDP of 140 trillion yuan. Even when including cloud computing and big data, the total reached only 1.6 trillion yuan, or about 1.2% of GDP.
Citi's micro-level survey aligns closely with these macro figures. Among the 1,800 respondents who had used an AI chatbot at least once in the past month, 83% used it for less than an hour per day, and only 5% exceeded two hours—clearly illustrating the "high penetration, shallow usage" pattern.
**Heavy User Behavior Signals Approaching Inflection Point** Citi believes that although overall usage depth remains low, several behavioral characteristics of heavy users point to an accelerating inflection point.
In terms of frequency, 89% of heavy users (those using AI for over two hours daily) engage with AI chatbots multiple times a day. In contrast, about half of light users (those with less than 30 minutes of daily usage) do not even use AI once daily. Regarding future intentions, 46% of heavy users plan to significantly increase their usage within the next year, compared to only 19% among light users.
Differences in usage scenarios are also noteworthy. Among heavy users, 64% utilize AI for work assistance, and 38% for creative thinking and brainstorming—applications that are far more complex than those typical of light users. Citi notes that heavy users also have significantly higher exposure to foreign AI products (such as ChatGPT), indicating earlier adoption and positioning them as typical early adopters.
These data points mutually reinforce Citi's assessment that AI adoption in China is evolving along a J-curve, approaching an upward inflection point.
**Preliminary Estimate: Economy-Wide Efficiency Gain of About 1.8%** As AI adoption is still in its early stages, Citi uses user "trust" in AI responses as a proxy indicator for efficiency improvement—higher trust corresponds to more time saved, while lower trust implies time wasted on verification.
Based on this framework, AI chatbots save respondents an average of 30.4 minutes per day, equivalent to 4.3% of their waking time. Combining this with a generative AI market penetration rate of 42.8%, Citi extrapolates this efficiency gain to the entire economy, estimating a potential efficiency improvement of approximately 1.8%.
Heavy users benefit more significantly. Calculations show that users with over two hours of daily usage can save more than two hours per day, about 20% of their waking time. Furthermore, 38% of heavy users believe AI already surpasses humans in reasoning and creativity, compared to only 13% among light users (less than 30 minutes daily). Citi suggests that as the virtuous cycle of increased usage, enhanced trust, and upgraded application scenarios strengthens, productivity gains will accelerate in a non-linear fashion.
**Dominance of Major Platforms, Specialization Key to Survival** Citi's survey indicates that China's AI chatbot market is dominated by leading internet platforms, but products are already showing clear differentiation in capabilities and competitive positioning.
Doubao from ByteDance ranks first with a 79% user penetration rate and is the "most frequently used product" for 63% of its users, achieving an NPS of 54.4. Tencent Yuanbao holds second place with 42% penetration and an NPS of 28.3. DeepSeek shows impressive performance with 37% penetration and an NPS of 36.1, particularly notable given its limited sales and marketing investments. Alibaba's Qwen and Baidu's Ernie follow with figures of 24%/15.3 and 23%/18.4, respectively.
Regarding key drivers for user choice, Doubao excels in speed and ease of use (recognized by 62% of its users each), Yuanbao wins on trust (57%), DeepSeek is appreciated for accuracy (66%) and logical reasoning (58%), while Tongyi leads in creative and emotional tasks (scoring 4.32 for creativity, the highest among the five products).
Citi believes that the diversification of AI applications will lead users to select different products based on specific needs, making it unlikely for a single product to become an "all-in-one AI gateway." This leaves room for platforms like Alibaba and Tencent to continuously expand AI applications within their respective ecosystems and suggests that future competition will intensify in vertical specialization and agent-based intelligence.
Additionally, 45% of respondents have used foreign AI products, with ChatGPT having the highest usage rate (30%), followed by Google's Gemini (13%) and Meta's Meta AI (8%). The primary reasons for using foreign products are seeking "different answers" (45%), better functionality (44%), and higher accuracy (43%). Penetration of foreign products is more prominent among users in higher-tier cities and iPhone users.
**Labor Impact Remains Largely Emotional, Formal Workplace Adoption Lags** Citi's prior macro-level analysis estimated that AI could affect approximately 31% of jobs in China, with about 9.6% (roughly 70 million jobs) facing direct replacement risk. However, the survey suggests that the tangible manifestation of these potential risks remains limited at this stage.
On an emotional level, 69% of respondents believe AI will replace a significant number of human jobs in the future, but only 40% express personal concern about their own jobs being affected—individual assessments are far more optimistic than overall judgments. 59% believe AI will help improve income opportunities, reflecting an expectation that complementary effects will dominate labor market sentiment. The overall level of optimism exceeds Citi's previous expectations.
Usage scenario data confirms that AI is still primarily used for personal purposes: general search (71%) and health consultations (51%) dominate, while work assistance (56%) and brainstorming (31%), though ranking high, show a clear lag in formal employer-led adoption. Only 6% of respondents reported that their employer or educational institution has designated specific AI tools for use; even among heavy users, this figure is only 12%.
Breaking down by age group, the 18-24 demographic shows the highest level of employment anxiety, with 58% worried their jobs could be replaced by AI—the highest among all age groups. Paradoxically, this group has the lowest depth of AI usage, with only 3% being heavy users (over two hours daily). Citi notes this aligns with its macro research conclusion—younger individuals and those in entry-level positions face higher exposure to AI, leading to more concentrated employment anxiety.
**Monetization Opportunities Emerge, Premium Features Outpace Content Subscription** The survey indicates that 58% of users prefer to continue using free services, consistent with historical patterns in internet industry monetization. However, Citi sees significant monetization potential revealed by the fact that 45% of respondents are willing to pay for advanced features.
Willingness to pay increases with age, city tier, and among iPhone users. Among those willing to pay, 63% cite "access to more advanced models" as the primary driver, followed by multimodal capabilities (document analysis, image generation, advanced voice interaction, 53%) and faster response times (51%).
The average acceptable monthly subscription price is 48.3 yuan, higher than typical digital content subscriptions like music and video, and roughly equivalent to the average monthly spending of online gamers. Citi believes this indicates that users applying AI to complex tasks have a stronger willingness to convert to paying customers, and that the reasonable pricing range for AI service subscriptions is higher than current market perceptions suggest.
Comments