Tracking Chinese AI search: a different web, different rules
Chinese AI surfaces are trained on a different web with different citation hierarchies. Probe data found Wikipedia presence was the strongest predictor of brand mention rate while on-site schema density was uncorrelated and mildly inverted — the opposite of standard Western AEO advice. Access, not parsing, is the hard part.
The second largest AI search market in the world
The Chinese ecosystem is not an edge case. Baidu ERNIE alone serves AI answers to hundreds of millions of people monthly, and it has evolved into something closer to AI Overviews than to a chatbot.
The signals invert
One probe of Chinese surfaces found Wikipedia presence was the strongest single predictor of brand mention rate, while on-site schema density was uncorrelated and mildly inverted. If your AEO playbook is structured data and content clusters, it will underperform there.
Where each one draws from
- Doubao pulls from Douyin, Toutiao and Xigua — the ByteDance universe, not the open web.
- Kimi is the long-context one, which is where B2B whitepapers actually get read.
- Baidu ERNIE leans heavily on Baike, Zhidao and Baijiahao.
- Qwen sits inside Alibaba's commerce and cloud surfaces.
Access is the moat
DeepSeek and Qwen gate chat behind accounts and Doubao wants a Chinese phone number. That difficulty is precisely why no general-purpose API sells this data and why the agencies who need it are currently maintaining their own scrapers.
We don't farm accounts
See this for your brand
Nostimates shows you how your brand shows up across twelve AI engines and search — in a dashboard, or as data in your own tools. Tell us what you want to measure.