Right now, a Chinese consumer is asking Doubao to recommend a foreign skincare brand. Your brand doesn’t appear. A competitor does. That consumer buys from them, not you. This is happening thousands of times a day, and most foreign beauty brands have no idea it’s going on. If you’re still optimizing for Google while your Chinese customers use AI assistants to find products, you’re not behind on tactics. You’re behind on reality.
What GEO Is and Why Chinese Consumers Have Already Moved On
GEO stands for Generative Engine Optimization. It means structuring your brand content so AI models, not search algorithms, recommend you in their answers. In China, those models are Doubao, Kimi, DeepSeek, and Baidu AI. They’re not fringe tools. They’re where product discovery happens in 2026.
According to iResearch’s 2026年GEO生成式引擎优化行业研究报告, more than 35% of product discovery queries in China now go through AI chat interfaces. That number was near zero three years ago. The shift happened fast, and most foreign brands missed it entirely.
When a user types “推荐一个好用的外国护肤品牌” (recommend a good foreign skincare brand) into Doubao, the AI pulls from structured product data, ingredient content, social proof signals, and brand entity consistency across platforms. If your brand doesn’t have that content infrastructure in Chinese, the AI ignores you. Not because of an algorithm change. Because you gave it nothing to work with.
| AI Assistant | Monthly Active Users (2026 est.) | Share of Beauty Queries |
|---|---|---|
| Doubao (ByteDance) | 500M+ | High (lifestyle-first model) |
| Kimi (Moonshot AI) | 200M+ | Medium-high (long-form answers) |
| Baidu AI Search | 600M+ | High (purchase intent queries) |
| DeepSeek | 150M+ | Growing (research-heavy users) |
The Brands Winning GEO in China Right Now
Look at what the winners have in common. It’s not budget. It’s content structure.
La Roche-Posay has thousands of XHS posts with ingredient breakdowns written in Chinese. Their product pages on Tmall explain each active ingredient, its percentage, and its skin benefit. They have a KOC review corpus built over years. When Kimi answers a question about sensitive skin care for foreigners, La Roche-Posay appears. Not by accident. Because their content is exactly what AI models need to form a recommendation.
CeraVe did the same thing with ceramide education. Hundreds of Chinese-language articles, videos, and reviews explaining what ceramides are and why they matter. The brand became the answer to ceramide questions across AI assistants. Their Chinese sales reflect it.
Drunk Elephant is the counterexample. Strong brand globally, weak Chinese content infrastructure. Ask Doubao about clean beauty for oily skin and Drunk Elephant rarely shows up. That’s a GEO failure, not a product failure.
The winning formula is clear: rich XHS content in Chinese, detailed product pages with ingredient explanations, a consistent brand entity across Tmall, XHS, Douyin, and Weibo, and a KOC review base that AI models can pull from as social proof. That’s it. Not complicated. Just work most brands haven’t done.
If you want your brand to appear in those AI answers, start with Douyin advertising and XHS content. Those two platforms feed the data AI assistants use most.
Marcus Zhan’s Take
“If your brand doesn’t appear when a Chinese AI answers a beauty question, you don’t exist for that consumer. Period.”
Marcus Zhan, Douyin & China Digital Strategist
I’ve had this conversation too many times. A brand comes to me after two years of bad results in China. They have a nice Tmall store, some KOL posts, a WeChat account nobody reads. I ask them: what happens when someone asks Doubao to recommend your product category? They don’t know. They’ve never tested it. That’s the problem. GEO isn’t a future trend to “monitor.” It’s the current reality of how Chinese consumers discover products. If you haven’t tested what AI assistants say about your brand in China, you’re operating blind.
The brands that win aren’t smarter than you. They just started earlier and they publish more. GEO rewards volume and structure. Fifty detailed XHS posts beat one expensive TVC every time when a Chinese AI decides who to recommend. Your competitors know this. They’re building their content infrastructure right now while you’re reading reports and scheduling another internal meeting. Stop deliberating. Start publishing.
4 Things Slow Brands Do Wrong
- They optimize for Google, not Doubao. Your SEO team is writing English meta descriptions while Chinese AI assistants have zero information about your brand in Mandarin. You’re solving the wrong problem in the wrong language for the wrong market.
- They treat XHS as optional. XHS content is the primary data source AI assistants use for beauty recommendations. If your brand has 12 posts on XHS in 2026, you’re invisible. Full stop. Competitors with 500 posts own those AI answers.
- They launch without a KOC strategy. AI models weight social proof. They pull real user reviews, real ingredient discussions, real before-and-after posts. If your KOC corpus is thin or fake, the AI skips you. Authentic micro-creator content at scale is not a nice-to-have. It’s the foundation of your GEO presence.
- They have no consistent brand entity. Your brand name is spelled three different ways across Tmall, XHS, and Douyin. Your product names don’t match. Your ingredient claims contradict each other. AI models can’t form a coherent recommendation from a brand that doesn’t know what it is. Fix your data hygiene before anything else.
5 Questions Brands Ask Me (And My Honest Answers)
Q: We already rank on Baidu. Isn’t that enough?
No. Baidu search and Baidu AI are two different systems. Ranking in traditional Baidu results does not mean Baidu AI recommends you. They pull from different content signals. You need to optimize for both, and most brands haven’t done either properly.
Q: How long does GEO take to show results?
Three to six months if you start now with serious content volume. If you’re publishing two posts a month, add another year. AI models need enough content to form a recommendation. Give them something to work with.
Q: Can we just translate our existing English content?
Translation is not localization. Chinese AI assistants expect Chinese-language content written for Chinese consumers, with Chinese skin concerns, Chinese ingredient vocabulary, and references to Chinese beauty culture. A translated press release doesn’t cut it.
Q: We have a big KOL. Does that help our GEO?
One big KOL does very little for GEO. AI models respond to volume and consistency, not one viral post. You need dozens of KOCs publishing regularly across XHS and Douyin. That’s what builds the content corpus AI assistants pull from.
Q: Is GEO the same thing we do for cosmetics marketing in China generally?
GEO is one layer of your China marketing. It works alongside your Douyin presence, your XHS strategy, your Tmall store. You can’t do GEO in isolation. But if you skip it, the rest of your China marketing is less effective because the AI discovery layer is broken.
Sources
- iResearch 《2026年GEO生成式引擎优化行业研究报告》 relay, 020yueshi.com
- iResearch 2025 China Digital Marketing Report, 100ec.cn
Your brand is either in the AI answer or it isn’t. There’s no middle ground in 2026. If you want to fix your GEO presence in China, start with your Douyin and XHS content strategy. Talk to us about Douyin advertising or contact us directly for a GEO audit of your current China presence.
