China’s New Micro-Drama Rules: AI Labels and a Live-Action Push

September 2026 brought two signals from Chinese regulators about short vertical dramas. On September 1, the country’s first dedicated rulebook for the format took effect, including a requirement that AI-made dramas carry a notice in every episode. On September 17, an official from the National Radio and Television Administration (NRTA) said live-action dramas are “the mainstay of quality productions and should receive strong support,” at a time when AI accounts for the large majority of new releases.

This post is about the rules and the policy direction. If you want the money side, how cheap AI production flooded the market and why so few titles earn back their costs, read our companion piece on China’s AI short drama boom. Here we stick to what the regulation says, what officials announced, and what it means for viewers and makers outside China.

The regulation is called the Administrative Measures for the Development of Micro-Dramas. According to the full text published by Xinhua, it was approved at an NRTA meeting on July 27, 2026, issued as Order No. 16 on July 30, and took effect on September 1, 2026. It covers dramas with episodes under 20 minutes that have a clear theme, a continuous storyline and prominent characters.

The Measures sort dramas into three classes. Class 1 covers large investments or special subjects such as politics, the military, diplomacy and national security; Class 2 covers smaller investments on general subjects; Class 3 covers low-investment general dramas. The text describes these in qualitative terms and leaves the detailed criteria to the NRTA. Class 1 dramas must be filed before production and receive a distribution permit; Class 2 dramas receive an approval document; Class 3 dramas are reviewed by the broadcasting platform, which assigns a program number. Penalties for breaches range from correction orders and public criticism up to fines of 100,000 yuan for some violations, and revocation of permits obtained by fraud.

Two articles get most of the attention abroad. Article 34 says that micro-dramas generated or produced with AI must follow national rules and carry a notice in a prominent position in every episode, and it puts that duty on both the production company and the platform that broadcasts the drama. In other words, a single disclaimer on a show’s landing page is not enough under the text; each episode has to say it.

Article 37 tells platforms to audit their recommendation algorithms regularly and bars algorithm models that induce users to become addicted or to overspend. TechNode summarized it as a ban on recommendation systems designed to drive excessive consumption or addiction. For a format built on cliffhangers and pay-to-unlock episodes, that is a notable line, though the text does not define where a compelling feed ends and an addictive one begins.

CGTN, reporting the September 17 briefing, added that micro-dramas must carry clear content labels, that a unified label has been introduced, and that regulators are working to make platforms follow rules on authorizing the use of people’s likenesses and voices. The Measures themselves do not spell out likeness rules, so that part rests on the officials’ statements.

The numbers explain why officials spoke up. CGTN, citing Xinhua, reported that China released 430,000 micro-dramas from January to August 2026, 13 times the total for all of 2025, and that AI-made productions made up more than 90% of them. Earlier figures show the share moving around: 36Kr, citing the China Netcasting Services Association, reported that more than 95% of first-quarter releases were AI dramas, while in the second quarter 153,000 of 239,000 releases were. TechNode put the first-half AI share at more than 74%. The figures cover different periods and counting methods, which is why they differ.

NRTA official Wang Xiaoliang said AI has cut costs and widened creative options, but that it has also created problems for content quality and for protecting rights. The response is support rather than a ban: according to CGTN, authorities will expand a program backing quality productions, guide platforms to create dedicated sections for live-action dramas, and offer financial support to help those shows reach audiences and earn returns. How large that support will be has not been published.

Chinese companies are a large part of the global short-drama business. CGTN quoted NRTA spokesperson He Biao as saying Chinese micro-dramas account for more than 80% of the overseas market, and officials said they intend to use AI to strengthen that position. The Measures support overseas distribution and say that dramas filmed in China for overseas broadcast are subject to the same filing and approval rules. Whether the per-episode AI notice will appear on dramas that Chinese-owned apps distribute abroad is not addressed in the coverage we read, and we would not assume it.

For viewers, the practical takeaway is that disclosure is becoming normal. China now requires an AI notice on each episode at home, and other jurisdictions are moving toward their own labeling rules; our guide to AI-generated video labeling rules tracks them. For makers, the lesson is to plan disclosure and consent records from the start, and to expect regulators everywhere to look harder at feeds engineered to keep people paying. This is a summary of public reporting and the published text, not legal advice.

ShortFling is a small platform outside China, and these rules are background for us rather than a direct obligation. Our vertical drama catalog lists 12 series, all marked coming soon with no playable episodes yet; you can leave a notification for launch. What is live today is reading and writing: the fiction library, with separate SFW and NSFW sections and the NSFW side behind an 18+ confirmation, and the writing studio, where authors draft with an AI editor that is free within a daily allowance and submit their stories for review before publication.