AI-driven commerce sellers now have to "show ID" starting today.

On August 10, a new WeChat Video rule took effect: if a product video uses AI in any form — a digital-human voiceover, an AI-generated product poster, or a synthetic demo of the product in use — it must be explicitly labeled. Miss the label and you face restricted distribution; traffic gets cut directly.

Per Jiemian News's August 3 report, the platform's "AI Application Rules for Short-Video Commerce" upgraded AI labeling from "recommended" to mandatory, listing four situations that must be labeled: AI-generated product display and usage footage, AI human presenters and digital-human appearances, AI-generated product posters and promotional images, and any other AI-generated video footage.

This time, the platform did something platforms hadn't done before: it split commerce scenarios from regular scenarios.

Regular Content Is Voluntary. Commerce Content Must Show ID

The rule's most elegant design is the split. AI creation in regular content is encouraged to be labeled — unlabeled, it still gets traffic. But once a video carries a product link, AI labeling shifts from moral obligation to entry requirement. No label: throttled. Labeled but mismatched with the actual product: also throttled.

The cut is precise. Over the past year, AI video generation got absurdly cheap: a few minutes per clip, mass digital-human voiceovers, AI-generated visuals, no human voice talent needed. Assembly-line content flooded the feed — AI faking real scenes to pose as genuine footage, digital humans impersonating real creators, deepfake narratives. Consumers increasingly can't tell whether a shot is real or synthetic, and honest creators can't outcompete the volume of low-cost AI trickery.

Moving the rule from "recommended" to "mandatory" is the platform's formal statement: AI content disguised as original human work is no longer tolerated.

Not Just WeChat Video: Every Platform Is Closing In

This cut didn't happen in isolation. Pull the timeline out and a tightening net comes into focus: on February 1, the Measures for Livestream E-commerce Supervision took effect, first establishing the legal requirement for conspicuous AI-content labeling; on July 15, the Interim Measures on AI Anthropomorphic Interactive Services landed, and Douyin, Kuaishou, and Bilibili tightened digital-human livestream reviews in tandem; on August 7, Xiaohongshu upgraded its AI governance rules, requiring AI virtual accounts to truthfully state their creation method in their profiles, with unlabeled AI content throttled — by May, Xiaohongshu had already processed over 180,000 AI-fraud notes and governed more than 1.2 million AI-managed accounts.

Overseas, the net is closing too: YouTube updated its policy against low-quality AI content last year, and Spotify purged roughly 75 million AI-generated tracks in violation. From livestream commerce to short-video selling, from notes to music, the "disguise era" of AI content is being turned over globally.

Many sellers fear that labeling "AI-generated" will hurt traffic. But the rule targets AI content disguised as original human work — compliant labeling won't throttle quality AI works; what gets hit is assembly-line junk riding on deception.With clear rules, low-quality account farms get flushed out gradually, and creators who actually understand products and content will win steadier traffic.

For AI-commerce sellers, the cheapest move is to make "AI label" a fixed checkbox in the publishing flow, rather than scrambling after the throttle. For ordinary users, at least when a labeled AI-generated commerce video shows up, you'll know the footage might not be real.

See you tomorrow.