China: AI content labelling rules
The rules on labelling AI-generated content in China, what each asks of a file, and what one Kennmarq.com call adds.
In short
- Images: Planned, January 2027
- Video: Planned, January 2027
- Audio: Planned, January 2027
- Text: Planned, March 2027
- 3D: Planned, June 2027
- Rules: 1 general, 3 for some sectors only.
- In the API:
markets=cn
Updated 4 November 2026; sources checked 4 November 2026. Sources are government sites. This page explains what Kennmarq.com does; it isn't legal advice.
Since 15 July 2026, companion (anthropomorphic) AI services must tell users they are talking to AI and remind over-reliant users that replies are AI-made.
General rules
- Labelling Measures and GB 45438-2025: In force since 1 September 2025. Visible or spoken labels with words like 'AI生成', and an 'AIGC' metadata label in GB 45438-2025's format.
Rules for specific sectors
Advertising and influencers
- Live-Streaming E-Commerce Measures (SAMR Order No. 117): In force since 1 February 2026. Label AI-generated people under national rules and keep reminding viewers they are AI-generated.
News and broadcast media
- Micro-Short Drama Measures (NRTA Order No. 16): In force since 1 September 2026. A notice label in a prominent position in every episode of a micro-drama generated or made with AI.
Platforms
- Algorithmic Recommendation Provisions, Art. 9: In force since 1 March 2022. Prominently label unlabelled algorithmically generated or synthetic content before passing it on.
In the API
A call with markets=cn is refused for now, before anything is counted: China's rules need China's metadata label (GB 45438-2025), expected January 2027; spoken notices in audio, expected December 2026.
China on the map, and every country's rules side by side.