2026 China Digital Human Influence Index Report

The 2026 “China Digital Human Influence Index Report” (2026中国数字人影响力指数报告), the fifth annual edition of the series carrying the English cover title China Metahuman Influence Index, was released by Shen Hao (沈浩), chief scientist of the State Key Laboratory of Media Convergence and Communication, at the AI Creator Economy Forum of the 13th China Network Audiovisual Conference in Chengdu in April 2026, according to the announcement by Communication University of China (中国传媒大学). The report was compiled by the State Key Laboratory together with the university’s Digital Human Research Institute (数字人研究院), and jointly released with the China Netcasting Services Association, the university’s School of Animation and Digital Arts, and Zhongke Trend Technology (中科趋势科技), which takes the co-publisher position held in the previous edition by Yuanli Trend Network. Following the conference debut, the full content was distributed in July 2026 through a wide-circulation release reviewing the whole of the 2025 industry year.

The report’s core judgment is that 2025 was the key inflection point in the industry’s development. Breakthroughs by large language models in multimodal interaction, autonomous planning, and tool invocation have restructured the sector’s underlying logic: the digital human is no longer a passive “display object” (展示物) but a “social actor” (社会行动者) able to perceive, decide, interact, and carry defined social functions, embedded in production, consumption, collaboration, and governance. For the first time the series organizes its analysis around three core propositions—agent evolution (智能体进化), digital assetization (数字资产化), and social embedding (社会嵌入). On industry scale, the report counts more than 2.23 million digital-human-related enterprises in 2025, with nearly 550,000 newly registered during the year, and describes digital humans as fully rooted in government services, culture and tourism, museums, and e-commerce livestreaming.

For the fifth consecutive year the report publishes ranking lists for the industry’s three core tracks, built from full-network data collection on 374 key digital human IPs across five platforms—Weibo, Douyin, Bilibili, Kuaishou, and Xiaohongshu—scored on the series’ three-dimensional influence index of communication power, product power, and social power. This edition adds three firsts: a technical-art grading system, a digital human asset evaluation framework, and a dedicated study of cultural digital humans. The 2025 virtual idol list (Top 11) comprises Luo Tianyi (洛天依), A-SOUL, Infinite Kings Troupe (无限王者团), Liu Yexi (柳夜熙), Xingtong (星瞳), Lingyuan Yousa (泠鸢Yousa), Tianyu (天妤), Li Zeyan (李泽言), Lili (厘里), Su Xiaomei (苏小妹), and the AI-native newcomer Yuri (尤栗); the report reads the track as crossing from motion-capture-driven to AI-native production, with culture-themed idols as the main force breaking into wider audiences. The virtual anchor list (Top 9) comprises Gu Xiaoyu (谷小雨), Ling Yanhuan (令颜欢), CCTV.com’s Xiao C (央视网小C), Hanser, Ai Xia (爱夏), Xiao Zheng (小诤), Nanako (菜菜子Nanako), Guan Xiaofang (关小芳), and the digital human of Luo Yonghao (罗永浩), with media-type and e-commerce-type anchors achieving monetization at scale. The digital employee list (Top 8) comprises the Haier Brothers (海尔兄弟), Liu Sanjie (刘三姐), Kedalin (科达琳), OPPO’s Xiaobu (OPPO小布), Congmei (葱妹), Xiaoyi (小意), Chengluoluo (橙络络), and Aijia (艾珈), a track the report describes as upgrading from basic execution to professional collaboration. Across the three lists, head IPs are characterized as balancing four dimensions of technical, communication, commercial, and social strength.

A newly added agent chapter (智能体篇) presents multimodal large models as rebuilding the digital human’s intelligent architecture, moving it from “human-like” toward “intelligence-like” through an integrated cognitive loop of perception, memory, reasoning, expression, and action. It details five capability leaps: native multimodal input with cross-modal alignment at the perception layer; million-token context plus vector databases for durable personalized memory; multimodal understanding, causal inference, and task planning at the reasoning layer; synchronized voice–motion–emotion output at the expression layer; and tool integration with vision-language navigation at the action layer, making the digital human an actual executor rather than a scripted presenter.

The digital assets chapter, the first systematic treatment of assetization in the series, likewise designates 2025 as the inflection year in which digital humans shift from “static credentials” to “dynamic subjects”—tradeable, programmable, derivable digital assets. It constructs a four-dimension asset evaluation matrix covering cognitive assets, relationship assets, commercial monetization assets, and technical/functional assets, and argues that digital humans are becoming bookable intangible or data assets in enterprise digitalization, while warning of guardrail problems including ownership disputes over agent-generated content, valuation difficulty, and deepfake reputation risk.

On the industry environment, the report records 28 provincial-level and 42 city-level units issuing nearly 300 special policies, a “standardization plus scale” dual-drive pattern it documents in a table of 2025 national measures ranging from the data infrastructure guideline of January 2025 through the AI-generated content labeling measures of March 2025 to the automotive and engineering-consulting policies of September 2025. It counts 165 investment and financing events in 2025, of which the AI Agent segment took 83, more than half, with capital logic shifting from single technology nodes toward integration depth, closed commercial loops, and cross-ecosystem synergy. Patent applications in the digital human field reached 737 for 2025, up 14.98% year on year, with invention patents at 98.5% of filings, an overall grant rate of 48.71%, more than 95% of patents deeply integrating AI, and emotional computing and reinforcement learning filings growing 35% and 28% respectively.

The cultural digital human study catalogues 81 representative figures—52 from traditional Chinese culture and 29 from revolutionary culture—spanning virtual idols such as Tianyu (天妤) and digital Su Xiaomei (苏小妹), virtual anchors such as Gu Xiaoyu (谷小雨) and the Henan broadcaster Tang Xiaomei (唐小妹), and institutional digital employees fielded by museums, publishers, and local governments. It diagnoses the segment’s difficulties as image homogenization, hollowed-out meaning, and a technology–emotion gap, and proposes three construction principles: respect the historical prototype, keep the image’s expression positive, and put technology in service of content. Cited exemplars include Zhonghua Book Company’s Su Dongpo digital human with more than five billion cumulative topic exposures, Huazhong University of Science and Technology’s Chuyin (楚音), described as the first cultural-relic IP digital human to take part in major current-affairs reporting, and Renmin University of China’s Zhao Yiman digital human.

In accompanying remarks, chief advisor Shen Hao states that the deep fusion of agent technology with the digital asset ecosystem is the industry’s next core direction, naming embodied intelligence, compliance, and emotionalization as the keywords for 2026. Digital Human Research Institute dean Lü Xin (吕欣) identifies human-like cognitive architecture—observation, planning, reflection—as the key breakthrough moving the industry from functional tool to value carrier, and research group leader Zhang Lijin (张丽锦) describes digital humans as becoming ubiquitous social units, with multimodal interaction and seamless cross-terminal migration expected as standard capabilities.