Kimi sources tracking
Kimi's source set skews toward long documents — whitepapers, filings, technical manuals — which makes it the clearest signal available on whether your B2B material is being read.
How it renders
A document-level reference set with passage grounding.
Sampling
Source panels are wider and more stable than inline citations, so they reach a usable confidence interval at lower n.
Gotcha
What the API returns
A single collection call against Kimi for the sources field, sampled n times and returned as one object per prompt.
{
"engine": "kimi",
"field": "sources",
"prompt": "best ai visibility tracking tool",
"country": "GB",
"language": "en",
"n": 30,
"sources": [
{
"url": "https://nostimates.com/docs",
"domain": "nostimates.com",
"inline": true,
"sample_frequency": 0.7
},
{
"url": "https://example.org/report",
"domain": "example.org",
"inline": false,
"sample_frequency": 0.23
}
]
}Response fields
| Key | Type | Notes |
|---|---|---|
| url | string | Source URL from the panel. |
| domain | string | Registrable domain. |
| title | string | null | Panel title as rendered. |
| inline | boolean | Whether this source was also cited inline. |
| sample_frequency | float | Share of n samples containing this source. |
What it costs
Included in the base collection wherever the engine exposes a panel.
See credit pricing →Kimi sources FAQ
- Does content clustering help on Kimi?
- Less than depth does. One thorough document tends to beat ten thin ones.
- Do PDFs need to be accessible?
- Yes — PDF accessibility matters more here than site structure.
See whether your whitepapers are being read
We will run your prompts against Kimi and report document-level citation for your library.
Other Kimi data
Across every engine
Compare how sources behaves on every engine we track.
Sources across all engines →