---
title: "China's AI Labeling Measures Put Visible and Invisible Marks on Generated Content"
date: 2026-11-22
category: Policy & Governance
site: NeuroAI
canonical: https://neuroai.site/a/na-policy-ai-content-labeling-measures
language: en
---

# China's AI Labeling Measures Put Visible and Invisible Marks on Generated Content

> A four-department measure and a companion mandatory standard set out how AI-generated text, images, audio and video carry labels in China.

A short video ends with a small line of text: AI-generated. A news image carries a mark in its file that no viewer sees. Both traces come from the same rulebook, now in force across Chinese platforms.

## Key takeaways

- The Measures for Labeling AI-Generated Synthetic Content (人工智能生成合成内容标识办法) were issued by four bodies — the Cyberspace Administration of China, the Ministry of Industry and Information Technology, the Ministry of Public Security and the National Radio and Television Administration — on 14 March 2025.

- The measures took effect on 1 September 2025 after a six-month transition period, according to the official announcement on the CAC website.

- A companion mandatory national standard, Cybersecurity Technology — Methods for Labeling AI-Generated Synthetic Content (网络安全技术 人工智能生成合成内容标识方法), took effect on the same date.

- The text defines two label types: explicit labels (显式标识) that users can perceive, and implicit labels (隐式标识) embedded in file metadata.

## Document type and effective stage

This is a departmental measure, not a law and not a draft. The four issuing bodies published it on 14 March 2025 and set 1 September 2025 as the implementation date. The official notice describes a six-month interval between publication and effect.

The document builds on three earlier sets of rules: the Provisions on the Governance of the Online Information Algorithmic Recommendation Services (互联网信息服务算法推荐管理规定), the Provisions on the Governance of Deep Synthesis of Online Information Services (互联网信息服务深度合成管理规定), and the Interim Measures for the Management of Generative AI Services (生成式人工智能服务管理暂行办法). Those earlier texts had already raised labeling questions; this measure sets out the implementation format.

## The two-label mechanism

The measure covers synthetic text, images, audio, video and virtual scenes produced with AI tools.

**Explicit labels** are added to the content or the interface and presented through text, sound or graphics in a way users can perceive. For text, state media reporting describes marks placed at the start, end or a prominent position, using wording such as "AI-generated" (AI生成) or "generated by artificial intelligence" (人工智能生成).

**Implicit labels** are inserted into file data through technical means and are not readily perceptible. State media reporting on the companion standard says the metadata field name contains the marker "AIGC" and can record generation status, the service provider, the distribution platform, a unique identifier and a digital signature or hash value.

The two layers serve different functions. One is visible to a reader; the other is designed for tracing a file after it has been copied.

## How duties are distributed

The measure assigns different tasks to different parties:

- **Generative service providers** add explicit labels to generated content, ensure downloaded, copied or exported files retain them, and insert implicit labels in file metadata.

- **App distribution platforms** ask applicants whether their apps provide generative AI services during listing review, and check labeling materials.

- **Users posting generated content** through online distribution services declare it and use the labeling tools the provider offers.

- **All organisations and individuals** are barred from maliciously deleting, altering, forging or hiding labels, and from supplying tools or services for those acts.

The scope is defined by the service, not by the technology alone. Organisations and research bodies that develop or apply generative AI without offering it to the public in China are outside the measures' stated scope.

## The standard published alongside it

The companion document is a mandatory national standard, published at the same time and effective on 1 September 2025. Its stated function is to give technical form to the labeling obligations in the measure. Chinese standards-body officials quoted in state media described the pairing as a combination of a management rule and a technical standard.

A figure from a separate report gives a sense of scale. A CNNIC report cited by state media put China's generative-AI user count at 515 million as of June 2025 — a population large enough that a label appearing on a feed is no longer a niche interface detail.

## How platforms have responded

State media reported in October 2025 that some short-video viewers had begun seeing AI-generation prompts at the bottom of clips. Platforms and service providers have had to build label insertion into export paths and metadata fields, and app stores have had to add a declaration step to review workflows.

These are observable implementation responses. The measure itself sets out the framework; the engineering work of applying it sits with companies.

## Honest limitations

The description here follows the official CAC notice, the government website explainer and state media reports on the companion standard. I have not reproduced the full text of either document, so the metadata field list reflects reporting rather than a clause-by-clause reading of the standard. The 515 million user figure comes from a CNNIC report as cited by state media, not from a directly accessed copy. Implementation practices vary by platform and are not surveyed here.

## What readers can do now

- Check whether an exported AI-generated file still carries its label after editing; losing the mark is a practical issue regardless of the rule.

- If you distribute generated content, use the platform's declaration tool rather than relying on a label added at creation.

- For compliance work, read the measure and the mandatory standard together — the first defines obligations and the second defines the technical format.

---

Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-policy-ai-content-labeling-measures
Free to quote with attribution and a link to the original.
