Xiaohongshu Now Requires AI Avatar Accounts to Declare Themselves

On August 7, 2026, Xiaohongshu (小红书) widened its AI disclosure rule. Until now, the platform only forced individual posts, 笔记, that were generated or heavily edited by AI to carry a label. From that date, the rule reaches the account itself. Any account run as an “AI虚拟人”, an AI virtual persona, must say so honestly on its profile page and turn on a dedicated “AI虚拟人” badge.

The mechanic is simple, according to a detailed report by Sohu News. A creator opens the profile edit page, goes to the profession field, and picks “AI creator, AI虚拟人”. Xiaohongshu says the declaration itself does not hurt reach: “主动标识不影响流量”, proactive labeling does not affect traffic. Accounts that disclose early even get priority access to new AI creation tools. Accounts that don’t, and get caught, face graduated penalties: posts buried, recommendations pushed to zero, eventually account-wide traffic restriction. Other users can report an account they suspect is undeclared.

Financial outlet Eastmoney, citing Shanghai Securities News, confirmed the same date and the same wording. This follows the “AI Generated and Synthetic Content Identification Measures” that took effect in September 2025 and already tag more than 300,000 posts a day. Xiaohongshu has banned 420,000 accounts for violations this year, and audience data cited in Chinese coverage shows 73% of users swipe away from purely AI content within three seconds. The platform has a real incentive to keep travel content, one of its biggest categories, feeling human.

What this actually changes

The note-level rule was already a year old. Nothing new there. What’s new is the account level. Xiaohongshu used to ask “is this specific post AI?” Now it asks “is this whole creator AI?” That’s a different question, and a harder one for the platform to dodge with a technicality.

I think this will strengthen trust in travel content on Xiaohongshu, not weaken it. Here’s why. Travel decisions are expensive and emotional. Nobody books a 12,000 RMB trip to Iceland off a note they suspect is fake. The 73% three-second swipe rate on pure AI content is not a curiosity, it’s the market telling the platform what it thinks of synthetic travel advice. Xiaohongshu is protecting its own commerce funnel by forcing the label.

But I don’t think this kills AI-run accounts. It legalizes a two-tier market. One tier: declared AI虚拟人 accounts, still allowed to post, still eligible for traffic, just tagged. The other tier: everyone else, human or claiming to be human. Some readers will avoid the tagged tier on principle. Others won’t care, as long as the destination info checks out. Xiaohongshu isn’t betting AI content disappears. It’s betting that disclosure lets the algorithm and the reader sort it out, instead of the platform having to guess.

The part that stays blurry is the middle ground. A large share of travel KOC content today is what agencies call semi-automated: a real person picks the destination and the angle, AI writes the first draft, AI or a template builds the video, a human posts it under their own face and name. Is that an “AI虚拟人” account? Probably not, by the letter of the rule, since there’s a real person behind it. But the line between “AI-assisted human” and “AI persona wearing a human face” is exactly where most agencies, ours included, sit today. Xiaohongshu hasn’t published a clear test for that middle case. Expect a wave of appeals and inconsistent enforcement in the first few months, the same pattern we saw when the note-level rule launched in September 2025 and 600,000 low-quality AI notes got pulled within six months.

One more thing worth flagging. A branded AI avatar KOL called 方桃子 reportedly bills 168,000 RMB for a 20-second video. That’s a real commercial layer, not a side experiment. Xiaohongshu isn’t trying to kill that business. It’s trying to make sure the audience knows what it’s paying attention to.

What this means for your brand

If your travel or hospitality brand runs Xiaohongshu content through KOCs, this rule touches you whether you knew it or not. Two questions to ask this week: does any account posting about us run on an AI persona, and does our agency know the difference between “AI-assisted” and “AI虚拟人” the way Xiaohongshu now defines it?

We work with a boutique hotel group in Portugal, small enough that I can describe them without naming them. Two years ago they built their Xiaohongshu presence entirely on human KOCs, real guests and real local creators, no AI avatars, no templated video voices. It cost more per post than the semi-automated route their competitors used. Slower, too. This month, that choice became a selling point instead of a cost. Their account manager is adding “100% verified human creators” to the next campaign brief, because it’s now something they can prove, not just claim.

Not every brand needs to make that pivot. A budget tour operator selling volume bookings cares less about a badge than about reach and cost per click. But for anyone selling trust, luxury stays, small-group tours, destination weddings, medical travel, the badge question is going to come up in client conversations before the end of the year. The window to get ahead of it is short. Once “verified human” becomes a standard claim on Xiaohongshu, being late to make it stops being a differentiator and starts being a gap.

How we help brands respond

The first step is not creative, it’s an audit. We go through a brand’s existing Xiaohongshu presence, every KOC, every branded account, every collaboration post, and check three things: who is actually posting, whether any tool in the content pipeline crosses the 30% AI-contribution threshold that triggers auto-labeling, and whether the account itself risks being read as an AI虚拟人 under the new rule. Most brands have never mapped this. They know their creators by name and follower count, not by production method.

From there the path splits in two, and we help the brand pick one deliberately instead of drifting into it.

Some brands go compliant and quiet: any account that touches AI generation gets the label turned on, cleanly, before the platform flags it for them. That protects reach and avoids the graduated penalties Xiaohongshu describes: buried posts, zero recommendation, account demotion.

Other brands go the other way and lean into “100% human verified” as a positioning line, the way our Portugal client did. That means tightening the KOC roster to creators we can vouch for personally, documenting it, and saying so in the content itself, not just in a pitch deck. We build this kind of vetted creator network as part of the Xiaohongshu and KOC strategy we set up for hospitality clients, drawing on the same approach detailed in our Xiaohongshu marketing guide.

Either path needs the audit first. We already saw what happens to accounts caught by surprise when the platform went after fake tourism accounts earlier this year, and how differently brands with clean rosters came out of it. For brands still building trust the slower way, the 听劝 model we wrote about, travelers asking real strangers for real advice, is exactly the kind of content this new rule was written to protect.

One question for the room

Some agencies will read this rule as proof that “verified human” is now worth charging more for. Others think Xiaohongshu just made disclosed AI content respectable enough that the price gap between human and AI-assisted creators will shrink, not grow. Which way do you think the market moves, and is your current KOC roster ready either way? Tell me in the comments, or write to us directly.

---