A China Internet Network Information Center (CNNIC) report put the country’s generative-AI user base above 700 million by the end of June 2026 and said penetration had moved above 50%. The published threshold is the central result: the public accounts do not state how far beyond 700 million the count reached. South China Morning Post reported that the figure covered more than half of the population, while China Daily described the report’s penetration rate as exceeding 50%.
The midyear marker comes after 602 million generative-AI users and 42.8% adoption were reported for December 2025 in CNNIC data carried by the State Council’s English-language site. SCMP characterized the end-June user base as 16% above that end-2025 level. Because the newer number is only published as a figure above 700 million, it supports neither a precise June total nor a separately calculated growth rate.
CNNIC reports a new midyear adoption threshold
CNNIC released 《生成式人工智能应用发展报告（2026）》, descriptively rendered as Generative Artificial Intelligence Application Development Report (2026), on September 29. Securities Times, citing CCTV News, said the CNNIC Policy and International Cooperation Institute released it at the 2026 (seventh) China Internet Infrastructure Resources Conference. Xinhua said the conference took place in Beijing. The accounts describe the report as using the latest survey data to examine user usage, industrial application, and policy formulation.
For this story, the usable public findings are tightly defined. The report places generative-AI users above 700 million by the end of June and its penetration rate above 50% in the first half of 2026. Those are CNNIC-attributed thresholds, not an exact user count. The source accounts also do not define, in the passages available, the full denominator used for the penetration rate.
That distinction matters when setting the mid-2026 measure beside the December 2025 baseline. The earlier CNNIC data give both a user total, 602 million, and an adoption rate, 42.8%. SCMP’s 16% comparison supplies a period-to-period description, but the threshold wording for June leaves the exact count undisclosed. The record is therefore strongest as evidence that the reported base crossed two visible markers, 700 million users and 50% penetration, within the stated period.
CNNIC’s reported snapshot centers on how respondents said they used generative AI. It offers no count of applications, ranking of products, or tally of daily users. That scope defines what the headline total measures.
Question answering leads reported generative-AI use
Question answering was the leading use identified in the report’s published summaries. The report found that 76% of users had used generative AI to obtain answers to questions. Xinhua characterized conversational use as the main application scenario. The figure identifies a reported use case among users; it does not establish time spent, a daily-use pattern, or which product those respondents selected.
China Daily’s account gives more detail on other reported tasks. It lists 47.8% for processing images or video, 37.6% for text-related tasks, and 32.5% for work summaries, meeting minutes, or presentation slides. Taken as the report’s survey results, those figures show that the published use-case picture extends beyond question answering while retaining the same user-based framing. They should not be treated as market shares or as counts of distinct products.
For related archive context, EastFrontier has covered China’s generative-AI app growth and Qwen office agents. Those are separate product and market stories, not evidence for CNNIC’s survey totals.
The work-oriented category groups work summaries, meeting minutes, and presentation slides together in its 32.5% figure. The report also says that use of general-purpose AI assistants and AI productivity tools each more than doubled from a year earlier, according to China Daily, without supplying underlying counts, category definitions, or a more precise time window.
A separate smart-hardware statistic requires equally careful reading. China Daily reported that 38.7% of Chinese internet users had bought smart hardware online in the preceding six months, including 20.2% for wearables and 18.2% for 3C products such as smartphones and tablets. That base is Chinese internet users, not generative-AI users, so the purchases cannot be recast as an adoption measure for generative AI.
The data points that the public report does not detail
The public accounts leave several key measurement details unstated. They do not provide a sample size, fieldwork dates, survey mode, weighting approach, respondent definition, margin of error, or complete definition of a generative-AI user. They say the report used the latest survey data, but that description alone does not establish how the figures were collected or how closely the reported group maps to any broader population.
The absence of an exact June figure is another practical boundary. More than 700 million is not interchangeable with 700 million, and the 16% comparison reported by SCMP should remain attached to its end-2025 reference point. The December baseline of 602 million and 42.8% adoption provides context, while the midyear disclosure establishes only that the subsequent count had passed the stated threshold.
The use-case percentages have their own limits. The 76% question-answering figure and China Daily’s 47.8%, 37.6%, and 32.5% figures describe what users reported doing; they do not reveal frequency, duration, spending, or the share of users attached to a particular service. Likewise, the finding that general-purpose assistants and productivity tools each more than doubled from a year earlier is not accompanied by raw totals or definitions that would allow a fuller comparison.