Skip to main content
Chainshift collects answers from the models below.
  • Requires agreement: available only to organizations granted access under a separate agreement. Without access, the model does not appear in the supported models list, and run requests are rejected with 403 (MODEL_NOT_GRANTED).
  • Location: whether the model supports city-level collection (location), which is narrower than a region. See Supported locations.
For behavior that differs per model, click the model name in the table to open its model-specific page.

Supported regions

The region of a run request is an ISO 3166-1 alpha-2 code. If you omit it, it defaults to KR. The 6 models other than the Naver family support the following 10 regions.
If even one model and region combination is unsupported, the entire run request is rejected with 400 (REGION_NOT_SUPPORTED_FOR_MODEL). Chainshift does not run only the supported combinations.
A region code not in the list above is rejected with 400 (REGION_NOT_SUPPORTED). Both errors are documented in Error handling.
Supported regions may grow over time. Retrieve the latest list with List regions. Add the model query parameter to return only the regions that model supports.

Supported locations

To collect at a narrower scope than a region (country), set location in the run request. The value is a city code under a country (for example, US-NEW_YORK). If you omit it, collection happens at the region level.
  • Currently only ChatGPT and Perplexity support location. If you include any other model in models, the entire request is rejected with 400 (LOCATION_NOT_SUPPORTED_FOR_MODEL).
  • If you also set region, the location must belong to that region. Otherwise you get 400 (LOCATION_REGION_MISMATCH). If you omit region, the run uses the region the location belongs to.
  • A location code not in the list is rejected with 400 (LOCATION_NOT_SUPPORTED).
Supported locations may grow over time. Retrieve the latest list with List locations. Use the model query parameter to filter locations by model, and region to filter by region.

What is fan-out

When generating an answer, an AI sometimes issues several search queries internally for a single user question, explores the web, and then synthesizes the results. The list of search queries it actually issued is called fan-out. Because these are search terms the AI created on its own rather than words a person typed into a search box, they are a useful signal for the search intents your brand is missing and the content topics you should create. The example below shows how one prompt becomes an answer by way of fan-out. One user prompt turning into three search queries the AI runs on the web before composing the answer One user prompt turning into three search queries the AI runs on the web before composing the answer Models that provide fan-out return this list of queries in the fanouts field of each answer. Models that do not provide it still generate the answer itself but do not expose fan-out queries, so fanouts is always an empty array. See the table above for which models provide it.
To learn how to use fan-out for content gap analysis, see the Usage guide.