They have no taste buds but they talk about "deliciousness"; they have no skin but they recommend "comfort"
Is the product endorsed by an AI virtual human “personally tested and effective” tenable?
Experts say that virtual humans cannot achieve "full use" within the meaning of the advertising law, and it is necessary to further clarify who is responsible.
Report from our newspaper (Reporter Chen Dandan) "Today I ate matcha layer cake, which tasted so fragrant and rich." "The spicy ghost pepper noodles were really spicy and fragrant." Recently, a reporter from "Worker Daily" searched for "AI virtual human" on some social platforms and found that some accounts have launched AI virtual human eating and evaluation videos, and marked "Can be opened for unboxing, evaluation, and immersive eating and broadcasting" on the homepage.
The reporter noticed that AI virtual humans introduced products in some promotional videos, and the products were not simultaneously released with real-life evaluation videos or copywriting. In a mukbang video posted by an account, AI virtual humans evaluate snowskin mooncakes, snail noodles, turkey noodles and other foods, and related product display screens are also generated by AI. In the window of this account, there are purchase links for similar products.
Not long ago, in a commercial video released by the producer of an AI short drama, the AI virtual person said, "My contact lenses are very comfortable after wearing them for a whole day." Whether the AI virtual person can "personally test the taste" and "personally test whether it is effective" has attracted attention. The reporter learned that at present, the products endorsed by AI virtual humans include medical devices, health care products, clothing, food, etc. Some AI virtual humans publish personalized feelings in the first person.
The reporter downloaded the AI virtual human video production software and tested it and found that the software can randomly generate the spoken copy of the AI virtual human. The membership fee for some software is 28 yuan per week. Related introductions say that after opening the membership, AI virtual human delivery videos can be quickly and batch generated.
Regarding the controversy caused by AI virtual human endorsements, Zhu Wei, associate professor at China University of Political Science and Law, said that according to the relevant provisions of the advertising law, the spokesperson must have used the product or received the service himself, or a close relative has fully used it before he can make a recommendation or prove it. However, AI virtual human beings cannot complete "full use" in the sense of the advertising law. He also emphasized that taking contact lenses as an example, the review standards and procedures for medical device advertisements are different from ordinary advertisements, and relevant supervision cannot be circumvented just because AI virtual human endorsements are used.
In September this year, the State Administration for Market Regulation released a number of typical cases of false propaganda in online food sales. In one of the cases, the party concerned used an AI virtual person in the live broadcast room to promote the health effects of ordinary food such as "helping to improve immunity", which violated the relevant provisions of the Anti-Unfair Competition Law. This case is the first typical case investigated and dealt with by Beijing after the implementation of the "Measures for the Supervision and Administration of Live Streaming E-commerce" in which AI virtual humans falsely promoted the health-care functions of ordinary food through live broadcasts. It made it clear that AI virtual humans are not a "shield" for false propaganda.
Zhu Wei further analyzed that there is a type of virtual image that corresponds to a real natural person, such as the virtual image of some celebrities, identifiable voices, portraits or cartoon images, etc. It is not of course impossible to use them for endorsements.
“What really needs to be resolved is who is the responsible subject.” Zhu Wei believes that the responsible subject for the AI virtual human endorsement should be further clarified, such as whether the developers and users behind the AI need to bear joint and several liability.
When it comes to the division of responsibilities, Zhu Wei said that for AI virtual human endorsements that do not correspond to real people, in principle, the advertising publisher must be held responsible for ensuring the authenticity. If the advertising publisher cannot bear relevant responsibilities, the advertiser and advertising operator should bear joint and several liability.
"Worker Daily" (October 12, 2026, Page 04)
