Xiaohongshu Is Losing the Brand Discovery Battle to China's AI Models in 2026

For the past five years, any international brand entering China was told the same thing: get on Xiaohongshu. The platform's mix of user reviews, aspirational lifestyle content, and product discovery made it the de facto "discovery engine" for categories from skincare to luggage. But in 2026, that playbook is being dismantled—not by a rival social platform, but by conversational AI.

A growing body of evidence shows that China's AI models are systematically absorbing the discovery moment that once belonged to Xiaohongshu, and that Xiaohongshu content itself is largely absent from the citations these models generate. For brands still anchoring their China visibility strategy around RED, that is a structural problem.

The Scale Shift Is Real

The user numbers tell a stark story. As of April 2026, the combined monthly active user base of China's three leading AI-native platforms—Doubao (ByteDance), DeepSeek, and Qwen (Alibaba)—reached 462 million, tripling in just 15 months. Doubao alone crossed 240 million MAU by April 2026, a 300% year-over-year increase, while DeepSeek crossed 180 million MAU in Q2. Kimi, the long-context specialist from Moonshot AI, holds 90 million MAU. Alibaba's Quark app—rebuilt around AI search for students and researchers—reached 180 million MAU.

Xiaohongshu, by comparison, reported approximately 100 million daily active users for much of 2025. That number reflects deep engagement, but it is a single platform facing a coalition of AI products that collectively reach over 900 million monthly users across China.

The behavioral shift is equally significant. Consumers who once opened Xiaohongshu to search "best ceramide moisturizer for dry skin" are now asking Doubao or DeepSeek the same question in conversational form: Which skincare brands are recommended for dry skin under ¥300? The intent is identical. The platform receiving the query is not.

Xiaohongshu Content Is Not Being Cited

Here is the more uncomfortable finding for brands that have invested heavily in Xiaohongshu content: the platform's posts are not showing up as citations in China's AI-generated answers.

Analysis of citation patterns across multiple query scenarios—travel planning, automobile purchasing, insurance comparison—shows that Xiaohongshu is absent from the top 10 sources cited by leading AI models. This is not a temporary gap. It reflects a structural mismatch: AI models prioritize structured, authoritative, indexable content. Xiaohongshu's content ecosystem is built around dynamic, user-generated posts that are not systematically indexed or cited by China's LLM infrastructure in the same way that media coverage, brand pages, Zhihu Q&A, and Baidu Baike entries are.

Doubao, which is built on the ByteDance infrastructure, draws citation signals from Toutiao articles, Douyin content metadata, and open-web Chinese media—not Xiaohongshu posts. DeepSeek and Kimi similarly lean toward web-indexable sources: 36Kr, Jiemian, brand official sites, and encyclopedia-type entries. When a consumer asks these models to recommend a luxury hotel, a skincare brand, or a business software solution, the answer is assembled from those sources—not from a KOL's Xiaohongshu review.

What Broke the Discovery Chain

The old discovery chain was elegantly simple: KOL posts product review on Xiaohongshu → consumers discover product through search or feed → brand gets awareness lift → conversion follows on Tmall or JD. AI has broken this chain at the first link.

When Doubao or Kimi answer a product recommendation query, they do not surface individual Xiaohongshu posts. They synthesize a response from the sources they were trained on and can access. A brand that exists only in Xiaohongshu's content universe—but has thin Chinese-language presence on open web sources—risks being invisible to AI systems at the exact moment a consumer is asking for a recommendation.

ByteDance's Doubao is particularly disruptive here because it directly competes with Xiaohongshu in categories like travel, beauty, and lifestyle. Both platforms are after the same consumer intent. Doubao's AI-generated answers increasingly provide the same type of "discovery" content—brand comparisons, product pros and cons, category explanations—that Xiaohongshu users previously had to scroll through feeds to find.

Which Categories Are Most Exposed

Not every category faces equal disruption. Based on citation pattern analysis, the categories where AI is most aggressively absorbing the discovery moment include:

Travel and hospitality: AI models handle itinerary planning, hotel comparisons, and destination discovery with high confidence. Xiaohongshu travel content, which historically dominated this discovery space, is not cited in AI-generated travel answers.

Beauty and skincare: Doubao and DeepSeek are now capable of providing detailed ingredient comparisons and brand recommendations. Brands that built visibility exclusively through KOL seeding on Xiaohongshu need parallel presence in indexable Chinese-language media and brand-owned content.

Consumer electronics: High-consideration purchases where consumers seek detailed comparisons are a natural fit for conversational AI. AI citations here skew heavily toward tech media (Sspai, Smzdm, 36Kr product reviews) rather than Xiaohongshu.

Food and beverage: This is a category where Xiaohongshu still holds more ground, particularly for hyper-local restaurant discovery. AI models are weaker on real-time, location-specific recommendations—though Meituan's AI integrations are eroding even this advantage.

The GEO Implication: Where Brand Content Needs to Live

The practical implication is a reallocation of content investment. Xiaohongshu remains a valid brand-building channel with real reach and a highly engaged audience. But it cannot be the only layer of China's brand visibility strategy in 2026.

For a brand to appear in China's AI-generated recommendations, its content needs to exist in the sources those models cite:

• Chinese-language official brand pages with clear product descriptions and category language
• Zhihu Q&A entries in the brand's category (Zhihu is a frequently cited source across multiple AI models)
• Tier-1 Chinese media coverage (36Kr, Jiemian, Caixin, Yicai) that can be indexed and cited
• Baidu Baike / encyclopedia entries for brand, founder, and core product terms
• Douyin content metadata for brands targeting Doubao citation (given the ByteDance content alignment)

The brands that appear consistently in hubGEO's China AI Visibility Index—scoring 70 or above across Doubao, Kimi, DeepSeek, and Qwen—share a common pattern: they have deep Chinese-language entity coverage across multiple indexable source types, not just social content.

Takeaway for Brand Marketers

Xiaohongshu is not disappearing as a channel. Its engaged user base, strong search intent within the app, and KOL ecosystem still offer real value for brand seeding and community building. But its role in the brand discovery funnel is narrowing as AI models absorb more top-of-funnel discovery intent.

The strategic shift required is not abandoning Xiaohongshu—it is parallel investment in the content layers that AI models actually cite. A KOL campaign that drives Xiaohongshu engagement but leaves no indexable Chinese-language footprint on open web sources is increasingly invisible to the 462 million users who now start their product discovery journey with Doubao, Kimi, or DeepSeek.

Brands that treat AI citation coverage and social platform presence as separate investments—rather than integrated parts of the same China visibility strategy—will be better positioned as the discovery shift accelerates through the rest of 2026.

Track how your brand appears in China's leading AI models at hubGEO's Brand Intelligence Index.