China Unveils AI Advertising Governance Framework to Tackle Fake Endorsements and Misleading Content
A new blue book recommends oversight across content creation, ad review, transactions and consumer protection while keeping advertisers accountable for their claims.
China's advertising industry is exploring a more comprehensive approach to regulating AI-generated promotional content, as increasingly sophisticated tools make it easier to produce realistic but potentially misleading images, videos, audio and endorsements.
The Blue Book on Internet Advertising Governance in the AIGC Era was released on October 24 during the 33rd China International Advertising Festival, according to China News Service reporting carried by Eastmoney. The document proposes a framework covering the advertising process from initial content creation and review to transactions and post-advertising consumer protection.
AIGC, or artificial intelligence-generated content, has rapidly expanded the range of tools available to advertisers. Businesses can use generative models to create product images, promotional videos, virtual presenters and personalised marketing materials at a fraction of the time traditionally required. These tools can lower production costs and help smaller businesses reach customers with more sophisticated campaigns.
The same capabilities, however, can also be used to produce deceptive material at scale. The blue book identifies risks including prohibited content, technical attempts to bypass moderation, problems in the transaction process and new advertising scenarios created by emerging technologies. It argues that traditional manual review alone may struggle to keep pace with the volume and speed of AI-generated material.
One of the document's central recommendations is to use AI systems to identify risks created by other AI systems. Multimodal models can analyse images, written text and audio together, potentially detecting misleading messages that might not be obvious when each element is reviewed separately.
For example, a promotional video may contain no clearly prohibited phrase when examined word by word, but the combination of its images, narration and implied claims may create a misleading impression. Context-aware analysis could help reviewers identify these cases more effectively than systems that search only for specific keywords.
Xiao Tao, a commercial-security content policy executive at Ocean Engine, said intelligent review systems could process 90% of advertising materials within 10 minutes at the fastest, improving efficiency by 75%, according to the China News Service report. Those figures describe the system's reported performance and should not be interpreted as proof that all violations can be detected automatically or that every reviewed advertisement is compliant.
The blue book also stresses the importance of maintaining clear responsibility. AI-generated content should remain subject to applicable laws and advertising standards, rather than creating a separate category in which businesses can avoid accountability. Advertisers are expected to remain responsible for the truthfulness of their claims, while platforms should maintain safeguards appropriate to their role and technological capabilities.
This distinction is particularly important for claims about health, financial products, consumer goods and endorsements by public figures. Realistic synthetic images and voices may make fabricated testimonials appear authentic. If such material spreads widely before being identified, consumers may make purchases based on false impressions about a product's effectiveness, origin or endorsement.
The proposed approach also extends beyond the advertisement itself. The blue book recommends stronger monitoring of transactions and the consumer experience after an advertisement generates a click or purchase. This reflects the fact that misleading marketing can involve more than a false image: a seller might advertise an attractive offer, direct users to another channel and then fail to deliver the promised product or service.
Platforms have begun introducing additional consumer-protection mechanisms. The report describes measures designed to make it easier for older consumers to submit evidence and reach human customer-service representatives, while parents can use standardised reporting processes to raise concerns involving children. Some cases may be resolved on the same day, although the availability and outcome of support will vary.
For businesses, stronger governance could raise compliance requirements but may also help protect consumer trust. Companies that use AI for marketing will need to consider how content is generated, whether product claims can be substantiated, how synthetic endorsements are presented and whether customer complaints can be addressed effectively.
The challenge is to balance innovation with enforcement. Overly rigid systems could delay legitimate campaigns or incorrectly flag harmless material, while weak safeguards could allow misleading advertisements to spread. Effective governance will therefore depend on reliable detection tools, transparent standards, human oversight and a clear process for appeals.
The release of the blue book does not by itself establish a new binding national law. It is a governance framework and set of recommendations rather than proof that every proposal has become a legal requirement. Its significance lies in the growing recognition that advertising oversight must adapt as generative AI changes both the speed and the methods of online promotion.