On 24 July 2025 the Cyberspace Administration of China launched the two-month Qinglang campaign against false information posted by self-media. What makes the notice distinctive is that it writes the labelling of AI-generated content directly into the enforcement requirements.
Four enforcement priorities
The campaign focuses on four problems: maliciously riding trending topics to mislead the public; distorting facts by various means; passing off content as genuine by failing to label it; and false information in professional fields.
The second category expressly identifies “using AI generation and synthesis technology to impersonate others and fabricate false social and livelihood information”. That turns AI from a tool into a named source of risk.
From labelling to admission to the recommendation pool
The notice sets two concrete requirements: improve source attribution and AI-generated content labelling; and information that should be labelled but is not may not enter the algorithm recommendation pool.
The second half is the key. It ties the labelling duty to traffic allocation — unlabelled content is not merely non-compliant, it directly loses the chance to be recommended. That design is more binding than an administrative penalty alone, because it reaches the platform’s commercial interest and content ecosystem.
It also dovetails with the design of the Measures for Labelling AI-Generated Synthetic Content. See our note on the Measures for Labelling AI-Generated Synthetic Content.
The enforcement chain of the labelling regime
From rule to implementation, the labelling regime has three links: the service provider adds labels at the generation stage; the distribution platform verifies and warns at the distribution stage; and the user bears a reasonable duty of care at the use stage.
This campaign concentrates on the distribution end, which suggests the regulator’s judgment that the generation end is largely under control while distribution remains the weak link. See our note on enforcement of synthetic content labelling.
Where platforms and content organisations can start
First, embed labelling verification in the publishing workflow, requiring uploaders to declare suspected AI-generated content.
Second, factor labelling status into recommendation decisions, so unlabelled content is not amplified by the algorithm.
Third, build fast handling and tracing for false information, preserving evidence along the generation and distribution chain. See our note on the reporting channel for AI application misuse.
The self-media ecosystem is a high-incidence zone for the spread of false information. Tying labelling to the recommendation mechanism moves governance from after-the-fact punishment to before-the-fact filtering. See our note on the internet information service measures for intelligent services.
This article is based on public information on the Cyberspace Administration of China’s Qinglang campaign against false information posted by self-media and is for reference only; it does not constitute legal advice.