China Names Fake Feng Shui Masters in a Short-Video Cleanup. An "AI-Generated" Label Won't Help

On September 30, the Cyberspace Administration of China, the country's internet regulator, announced a two-month campaign against fake personas on short-video platforms. It is part of the agency's long-running "Qinglang," or "clear and bright," series of cleanups. The notice lists five kinds of problems. Under the fourth, which it calls promoting extravagance and superstition, one item is unusually specific: accounts that dress themselves up as feng shui masters, astrologers or metaphysics mentors, push superstitious content, and use pitches about changing your luck and warding off misfortune to sell goods, charge for readings, or steer viewers into private chats where the selling continues. The same category covers rented-supercar wealth fantasies and rags-to-riches stories that end in a paid course; elsewhere the notice goes after accounts posing as delivery riders or struggling farmers to win sympathy.

The part that matters most to viewers comes in the enforcement measures. Platforms must make sure fake-persona videos carry the required content labels, and unlabeled ones will be dealt with severely. But the notice also says platforms may not let a video off because it is already tagged as fiction, AI-generated, marketing, a repost or personal opinion, or carries a disclaimer about having no harmful intent. A luck-changing guru who pitches a product at the end of the clip does not escape the campaign by adding "AI-generated" in the corner. The label describes how the video was made; the selling and the paid readings are still in scope. Another measure tells regulators to fold fields that lack an occupational category, qualification standard or certification, yet offer serious room for illegal profit, into routine enforcement, and bans fake-persona videos from trending lists and recommendation slots.

Platforms had already started. On September 22, Douyin, the Chinese version of TikTok, published a notice on content that lures users off the app. One of its four examples was fortune-telling: accounts styled as masters of Chinese classics or metaphysics hinted that they could change your luck, ward off disaster or rewrite your fate, then took the business elsewhere. Douyin said 821 accounts had recently lost livestream rights, been muted or been banned for posting such content or profiting from it. The other three examples were counterfeit cigarettes, illegal wildlife hunting and drug sales.

The notice mentions AI in two more places. Its third category targets impersonation "using AI and other means," including synthetic audio and video of public figures used for endorsements. Another measure singles out personas such as the "AI domineering CEO" and the "wellness master" used to deceive older people. On September 2, the same regulator reported on the second phase of a separate campaign against AI misuse: more than 5.61 million pieces of content removed, over 49,000 accounts penalized and more than 2,400 websites and apps dealt with. One of its published case types describes four accounts on Kuaishou, Xiaohongshu, Douyin and WeChat that used AI to fake disaster scenes and invent trending stories; the same case lists posting feudal superstition and building AI digital humans to sell products with false claims, misleading older viewers.

None of this names fortune-telling apps. The campaign is about short-video personas and the sales, fees and referrals built on them. Apps in China answer to a different rule: the Measures for Labeling AI-Generated Synthetic Content, in force since September 1, 2025. In the regulator's published Q&A on the rules, app stores must ask developers at review whether they offer AI generation and check their labeling. The September 2 report says Huawei, Xiaomi, OPPO and vivo app stores are spot-checking apps already listed. So an AI fortune app sold through Chinese app stores has to label its AI-generated content as the rules require. That label is a disclosure owed to the user. The new notice makes clear that on short video, the same kind of label is no shield.

For the next two months, the pattern to watch for is a master who tells you your luck is off, then offers something to fix it: a product, a paid reading, a private message where you send your birth date. That is the exact routine both the regulator and Douyin have named. Whether the clip says "AI-generated" and whether the master is a person or a digital human makes no difference: these accounts are being cleaned up, and paying them is a bad idea. The notice targets the persona-plus-sales package. It does not single out videos that simply talk about horoscopes without selling anything.