Kimi locale variance tracking
Kimi's variance is driven by document language more than by market. The same question against English and Chinese source libraries produces different documents and different conclusions.
How it renders
Controlled by prompt language, with document-language availability the dominant factor.
Sampling
Each locale is its own sample set. A cross-market comparison at n=30 per market costs what it says on the tin, which is why we surface the arithmetic before you commit.
Gotcha
What the API returns
A single collection call against Kimi for the locale variance field, sampled n times and returned as one object per prompt.
{
"engine": "kimi",
"field": "locale_variance",
"prompt": "best ai visibility tracking tool",
"country": "GB",
"language": "en",
"n": 30,
"locales": [
{
"country": "GB",
"language": "en",
"share": 0.34,
"divergence": 0.02,
"unique_domains": 4
},
{
"country": "DE",
"language": "de",
"share": 0.19,
"divergence": 0.15,
"unique_domains": 11
}
]
}Response fields
| Key | Type | Notes |
|---|---|---|
| country | string | ISO 3166-1 alpha-2 country of the egress used. |
| language | string | ISO 639-1 language the prompt was issued in. |
| share | float | Brand share within that locale. |
| divergence | float | Distance from the global mean for this prompt. |
| unique_domains | integer | Domains cited in this locale and nowhere else. |
What it costs
Each locale multiplies the collection count. Two markets at n=30 is 60 credits per prompt.
See credit pricing →Kimi locale variance FAQ
- Should I translate my whitepapers?
- If you want Chinese-language Kimi citation, yes. We will show you the gap first.
- Do you run both languages?
- Where it is useful, and we document the contrast.
See what translating your documents would buy you
We will run both language sets and quantify the citation gap before you invest.
Other Kimi data
Across every engine
Compare how locale variance behaves on every engine we track.
Locale variance across all engines →