On 23 April 2024 the Central Cyberspace Affairs Commission launched a Qinglang campaign against self-media accounts that pursue traffic without limits. The targets are manufacturing or hijacking trending topics, creating information traps, and attracting followers through prohibited means. The campaign also requires labelling of AI-generated information and of staged or fabricated content, and stricter conditions for opening monetisation permissions. For AI companies running content accounts, it turns labelling from a technical option into an operating precondition.
Three named behaviours
Manufacturing or hijacking a trending topic means faking the connection between content and event. An information trap means a misleading combination of headline, thumbnail and body. Prohibited follower acquisition means obtaining traffic through methods the rules forbid. What the three share is the exchange of information asymmetry for attention instead of providing real content.
For AI companies, the risk is not whether generative tools were used but whether generated material is presented as a real event. A batch of model-written “field reports” carrying real place names and timestamps is more likely to trigger action than obviously fictional content.
Where labelling lands
AI-generated information labelling and staged-content labelling are named together. The first addresses synthetic material, the second material staged to look like a real scene. Both answer the same question: does the user know what they are looking at.
A disclaimer is not enough. Position, prominence and wording all affect the outcome. A note in small type at the end differs materially from a clear statement at the start, and platforms are more likely to accept content that is clearly labelled.
Accounts and monetisation
Stricter conditions for monetisation mean that violation records affect an account’s ability to earn. For AI companies that use content for lead generation, an account is a marketing asset, and a single enforcement action can interrupt the whole acquisition chain.
The workable response is to move review upstream: scripts, assets, headlines and thumbnails pass a labelling and misleading-content check before publication, rather than waiting for a takedown and appealing. Appeals usually take longer than the content lifecycle, and by the time a decision arrives the campaign window has closed.
MCN and agency responsibility
Under signed-account and agency models, the agency has real control over content. The campaign brings agencies into scope, which means they need content management for contracted accounts: training, review and handling of violations.
Brands that delegate to an MCN should write content standards, labelling obligations and allocation of responsibility into the contract. A brand cannot fully disconnect by saying the agency produced the content; the account holder remains the party answerable externally.
Platform review and AI content
Platforms need detection and handling capability, including detection of synthetic content and treatment of unlabelled material. The labelling information a company submits can become the basis for platform decisions, and a false label is worse than a missing one.
For companies using AI-generated content in brand communications, writing labelling rules into the internal content handbook and keeping generation and human-review records is the cheapest form of protection.
For AI content labelling, see our AIGC labelling service; for infringement and takedown, see AI infringement response. Definitions are in the glossary.
This article is based on public information about the Qinglang campaign against self-media accounts pursuing traffic without limits, launched by the Central Cyberspace Affairs Commission on 23 April 2024. It is for reference only and does not constitute legal advice.