Kimi share of voice tracking
Kimi share of voice is normalised by answer length and reported with a 95% interval, so a brand mentioned once in a two-thousand-word answer is not scored like one mentioned once in a paragraph.
How it renders
Derived across the sample set with length normalisation applied.
Sampling
This is the field where n matters most. At n=1 the metric is noise; at n=30 the interval is typically tight enough to defend in a client report.
Gotcha
What the API returns
A single collection call against Kimi for the share of voice field, sampled n times and returned as one object per prompt.
{
"engine": "kimi",
"field": "share_of_voice",
"prompt": "best ai visibility tracking tool",
"country": "GB",
"language": "en",
"n": 30,
"share_of_voice": [
{
"brand": "Nostimates",
"share": 0.34,
"ci95": {
"low": 0.2,
"high": 0.52
},
"n": 30
},
{
"brand": "Competitor",
"share": 0.21,
"ci95": {
"low": 0.1,
"high": 0.38
},
"n": 30
}
]
}Response fields
| Key | Type | Notes |
|---|---|---|
| brand | string | Brand from the configured competitor set. |
| share | float | Point estimate across the sample set. |
| ci95 | object | 95% Wilson interval as { low, high }. |
| n | integer | Samples behind the estimate. |
| delta_prior | float | null | Change against the previous run of the same prompt set. |
What it costs
Derived at no extra credit cost, but it scales with n — the interval is bought with samples.
See credit pricing →Kimi share of voice FAQ
- Why normalise on Kimi?
- Because answer length varies by an order of magnitude between prompts. Raw counts are not comparable.
- What sample size?
- n=25 is often enough — Kimi is comparatively stable for a Chinese surface.
Report Kimi share without the verbosity bias
We will scope normalised share across your technical prompt set.
Other Kimi data
Across every engine
Compare how share of voice behaves on every engine we track.
Share of voice across all engines →