Sanbi China
Every plan runs Simplified-Chinese prompts on Chinese AI engines. An English-only audit measures the Western web, not the Chinese one — which is why each plan also runs the same prompts in English, so you can see the difference rather than guess at it.
China plans open soon — join the waitlist for yours and we'll email you the day it opens.
China Starter
Proof of whether Chinese AI engines know you at all. Five prompts across two AI search engines and both languages is twenty graded answers a cycle.
- Grounded in the Chinese web: Chinese prompts, Chinese AI engines, Chinese sources
- 5 prompts a month, each measured in Chinese and English
- On Qwen and DeepSeek
- 20 graded answers in every run
- Refreshed daily
- 1 domain
- 3 months of history
- Competitor tracking
- Chinese and English compared side by side
- Growth: the Chinese platforms worth being cited on
- 2 of 2 return sources you can act on.
China Professional
More prompts and more Chinese AI engines, for a team that tracks a whole category.
- Grounded in the Chinese web: Chinese prompts, Chinese AI engines, Chinese sources
- More prompts than China Starter, each measured in Chinese and English
- More Chinese AI engines: the ones that search the web and the ones that answer from memory
- Every answer graded, with the sources it cited
- Refreshed daily · 3 months of history
- Up to 5 domains
- Competitor tracking and brand aliases in both languages
- Growth: the Chinese platforms worth being cited on
China Enterprise
Prompt volume beyond the listed plans, quoted for your team.
- Grounded in the Chinese web: Chinese prompts, Chinese AI engines, Chinese sources
- Everything in China Professional
- Prompt volume set for your team
- Chinese and English compared side by side
- Unlimited domains
- A quote within a working day
Why Chinese prompts
The same forty prompts, translated into Simplified Chinese, moved the share of citations coming from the Chinese web from three or four per cent to around sixty, and roughly doubled the number of citations per answer. The serving region barely mattered; the language decided which web the model read. A Western tool asking in English cannot tell you anything about that, which is what this is for.