The beauty industry is one of China's most competitive consumer sectors, with international prestige brands, Japanese luxury skincare, and a rapidly growing domestic cosmetics scene all competing for consumer attention. Our tracking of 10 beauty brands across China's six major AI models reveals stark visibility divides that have direct strategic implications for every brand in the category.
The Perfect Score Pair: Estée Lauder and Lancôme
Only two beauty brands in our entire database achieve a perfect score of 100 across all six Chinese AI models: Estée Lauder and Lancôme. This consistency — across models with very different training methodologies, from Doubao's consumer-oriented approach to DeepSeek's open-source corpus — signals foundational brand authority, not platform-specific optimization.
Estée Lauder's dominance is anchored to one product: Advanced Night Repair (ANR), the "little brown bottle" that has become the most culturally embedded premium serum in China. ANR generates an extraordinary volume of Mandarin content: Xiaohongshu reviews, skincare routine videos, ingredient analysis articles, comparison guides, and dermatologist recommendations. This one hero product has given the entire Estée Lauder brand a Chinese AI training data footprint that spans years and multiple content formats.
Lancôme's parallel dominance is driven by the Génifique serum (小黑瓶 / "Little Black Bottle"), which has achieved similar cultural iconic status. The brand's L'Oréal Group ownership gives it access to deep China market infrastructure, consistent product launch coverage, and distributed media relationships that ensure continuous Mandarin text content production.
The hero product lesson: Both brands' AI dominance is disproportionately driven by single iconic products that generate far more content than the brand overall. Beauty brands without a clear hero product in China face a structural GEO disadvantage.
The Mixed Middle: SK-II, Shiseido, L'Oréal
Three brands show meaningful but uneven visibility across platforms, revealing different structural strengths:
SK-II (33/67/0/100/0/100): Strong on Qwen and Hunyuan, weak on DeepSeek and Wenxin. The Qwen and Hunyuan strength reflects deep e-commerce and WeChat ecosystem integration — Alibaba's Tmall and Tencent's WeChat are both core SK-II channels. The DeepSeek weakness suggests gaps in open-web academic and tech-adjacent content where SK-II's Pitera science story could be told more effectively.
Shiseido (67/0/0/0/100/0): The Wenxin perfect score suggests strong Baidu SEO investment, while zero on Kimi, DeepSeek, and Qwen indicates thin presence on social-commerce and open-source-weighted platforms. Shiseido's GEO position is Baidu-dependent — a single-platform risk.
L'Oréal (33/67/0/100/0/0): The corporate masterbrand is visible on Qwen (Alibaba ecosystem) and Kimi but absent elsewhere, likely because sub-brands (Lancôme, Maybelline) command stronger AI citations in product recommendation contexts.
The Near-Zero Tier: Clinique
Clinique scores a striking 100 on Qwen and zero everywhere else — the most concentrated visibility pattern in our entire dataset. This single-model dependence on Alibaba's Qwen suggests Clinique's AI presence is almost entirely driven by Tmall e-commerce data integration. The brand's open-web Mandarin content is insufficient to generate citations on models with more diverse training sources.
The single-platform risk: Any brand scoring high on one model and zero on others has a fragile GEO position. If Alibaba's training data integration changes, or if consumers shift toward non-Alibaba platforms for beauty research, Clinique's AI visibility would collapse.
The Zero Tier: Charlotte Tilbury, Drunk Elephant, NARS, MAC*
Four brands score zero across multiple models: Charlotte Tilbury (zero everywhere), Drunk Elephant (zero everywhere), NARS (zero everywhere), and MAC (zero except Kimi at 33).
These are not obscure brands — they have genuine Chinese consumer followings. The zero scores reflect a shared problem: all four brands have built their China presence primarily through social media visual content (Xiaohongshu tutorials, Douyin makeup looks) rather than AI-indexable text content. Beautiful as this content is, it doesn't train the text-based AI models that now mediate consumer discovery.
The visual content trap: Beauty is inherently visual, and social media rewards visual content. But China's AI models are trained primarily on text. Brands that focus exclusively on visual social media marketing are optimizing for human social engagement while neglecting the text-format authority content that generates AI citations.
The Beauty GEO Playbook
For brands at zero: Identify your single most credible claim and build 10 pieces of authoritative Mandarin text content around it. For NARS, that's "best blush for Asian skin tones." For Charlotte Tilbury, it's "Hollywood makeup secrets for Chinese faces." For Drunk Elephant, it's "biocompatible skincare ingredients explained."
For brands in the mixed middle: Diagnose platform-specific gaps. If you're strong on Qwen but weak on DeepSeek, you need more open-web scientific content. If strong on Wenxin but weak on Doubao, you need more ByteDance-ecosystem (Douyin, Toutiao) text content.
For brands already at 100: Maintain content velocity. AI training data has recency weighting — consistent new Mandarin content production is required to sustain high visibility scores as models are retrained.
The beauty category's AI visibility divide is ultimately a content strategy divide. Brands that treat Chinese AI models as a new distribution channel requiring their own content strategy will win. Brands that assume their social media momentum will automatically generate AI citations will find themselves invisible in an increasingly AI-mediated consumer discovery process.
Explore all 10 beauty brands' China AI scores at hubgeo.vercel.app/brands.