China Mandates AI Labels as AI Dramas Dominate Micro-Shorts

China’s media regulator will require every broadcast and audiovisual program that uses artificial intelligence to carry a clear label, ban unauthorized AI mashups of existing works, and keep human reviewers in the production chain, as new figures show AI-generated dramas now account for more than 90 percent of the country’s micro-short dramas.

The National Radio and Television Administration (NRTA) announced the measures at a State Council Information Office press conference in Beijing on September 17, part of the “Starting the 15th Five-Year Plan” briefing series on high-quality development of radio, television and online audiovisual services.

Labels, Human Review and a Ban on ‘AI Mogai’

NRTA Vice Director Liu Jianguo said programs generated, produced and broadcast using AI must add a corresponding label, that human review must be maintained at key production stages and for final-cut content, and that AI mashups and distortions are strictly prohibited.

“AI will keep evolving; the NRTA will keep improving rules,” Liu said, according to China News. The regulator added that it would make “scientific assessments and apply differentiated measures,” in the phrasing carried by The Paper, a Shanghai-based outlet citing CCTV News.

The anti-mashup provision targets a practice Chinese regulators call AI mogai—the disruptive alteration or vulgar adaptation of classic film, television and animation, particularly the Four Great Classical Novels, historical and revolutionary themes, heroic figures and children’s cartoons. A December 2025 NRTA campaign set three red lines for such content: violating the spirit or characters of an original work, adding blood, violence or vulgar material that transmits wrong values, and misappropriating Chinese culture by misaligning historical time and civilizational markers, according to an NRTA notice.

Liu framed the approach as deliberately balanced. “What should be opened up to encourage innovation will be; what must be guarded will be firmly guarded,” he said.

AI Now Dominates Micro-Drama Supply

The same briefing produced the statistic that has drawn the most attention: 430,000 micro-dramas were launched nationwide in the first eight months of 2026, 13 times last year’s full-year total, and more than 90 percent were AI dramas.

Wang Xiaoliang, director of the NRTA’s Online Audiovisual Program Management Department, presented the figures, Xinhua reported. The growth is the product of widespread, cheap AI adoption across animation, variety shows, web dramas and short-form video.

Micro-dramas—vertical, mobile-first series often under a few minutes per episode—have become one of China’s most-consumed entertainment formats. Average daily micro-drama usage reached 129 minutes per user as of December 2025, surpassing long-form video, and the market reportedly exceeded RMB 100 billion in 2025. AI has become the main engine of new supply: industry tracking cited by Yicai put the first half of 2026 at 367,000 releases, with AI-generated titles above 74 percent.

The gap between sheer volume and audience demand, however, is stark. Industry figures suggest only about one in 77 AI short dramas recoups its costs, meaning some 98.7 percent fail to break even. During the 2026 Spring Festival, live-action supply amounted to roughly one-fiftieth that of AI dramas, yet attracted about 25 times the total views. Quantity, in other words, is not the same as reach.

A Regulatory Framework Built Over a Year

The September 17 announcement caps a sequence of tightening measures. In November 2025 the NRTA launched dedicated governance of micro-dramas, folding AIGC and comic-style animated micro-dramas into tiered classification review. On January 1, 2026, a one-month nationwide campaign against AI mogai videos began. From April 1, AI micro-dramas were required to follow “register-before-launch” rules.

On September 1, NRTA Order No. 16—described as China’s first dedicated micro-drama regulation—took effect, requiring prominent AI labels in each episode and banning addiction-inducing recommendation algorithms. A week later, on September 7, the NRTA issued notices requiring that everything broadcast be reviewed and permitting review-while-broadcasting, while the Supreme People’s Court published opinions on AI-related disputes.

Enforcement has been visible. Platforms have run cleanups—one report counted nearly 700 titles removed in a single quarter—and several AI-generated dramas have been pulled for infringing the likenesses or works of real people. A Guangzhou Huangpu District court ruling held that AI-generated micro-dramas involving individualized selection and arrangement can be original works protected by copyright, on the principle that AI is a tool and people are the authors.

Analysis: Transparency as the Price of Scale

The new rules are best read as a transparency regime layered on top of an approval regime. The three-tier licensing system—key, ordinary and other micro-dramas—sorts titles by investment and subject matter, with sensitive topics such as politics, military affairs and national security routed to the strictest tier. The labeling and human-review mandates sit alongside it, and together they try to answer a question regulators elsewhere are also confronting: when machines produce most of the supply, who is accountable for what viewers see?

The core tension is speed. AI has cut production costs from tens of thousands of yuan per work to a few thousand, allowing four-person teams to produce hits on budgets that traditional studios cannot match. That efficiency is precisely what pushed supply 13-fold in under a year—and precisely why oversight, built through slow legislative steps, has struggled to keep pace.

Industry analysts argue the effect will be consolidation rather than contraction. TechNode has described the measures as moving China’s AI short-drama industry “from wild growth to tech-driven growth,” concentrating value among players with large intellectual-property libraries, foundational AI models and global distribution—firms that can absorb compliance costs that marginal operators cannot.

The mandate for human review also carries a labour implication. By insisting that AI cannot replace human gatekeeping at key stages, the NRTA is reserving a formal editorial role for people even as the craft of production is automated. Whether that creates meaningful oversight or becomes a nominal checkbox will depend on enforcement.

What to Watch

Three questions will define how these rules land. First, how the NRTA verifies labels at the scale of hundreds of thousands of titles—automated detection, platform self-reporting, or sampling. Second, whether the human-review requirement changes what gets made, since most AI dramas currently fail commercially and producers may respond by chasing licensed, higher-quality IP rather than volume. Third, how the framework interacts with copyright, given that roughly 70 percent of 2025 micro-drama works derived from online novels and courts are still defining where AI authorship boundaries lie.

China is not alone in this. U.S. micro-drama monthly active users climbed from 26 million in 2024 to a projected higher range in 2025, and the global short-drama economy is expanding quickly. China’s rules may become a template—and a test—for how a major market balances a booming AI entertainment sector against content integrity and public trust.

For now, the message from Beijing is consistent: innovation is welcome, but it must be labelled, reviewed and accountable. As Liu put it, AI will keep evolving, and so will the rules.