Skip to main content
Retrieve collected answers with GET /answers. You can call it at any time, regardless of whether the run has finished, and it returns the answers collected so far.

Retrieving answers

Below is a real collection response (the body is partially trimmed). The prompt is a Korean question asking for used SUV recommendations under 20M KRW.
This single response contains everything you need for AEO analysis: what ChatGPT answered (content), which pages it cited as evidence (sources, in this example two posts from a particular car blog), and which search queries it actually issued to build the answer (fanouts). Whether your site appears in the sources, and whether your content ranks for the fan-out queries, become your AEO action items.

Filters

Pagination

All list endpoints use cursor pagination.
  1. Send the first request without a cursor. Set the page size with limit (answers default to 20, maximum 100).
  2. Pass the nextCursor value from the response as the cursor parameter of the next request, unchanged.
  3. When nextCursor is null, you have reached the last page.
If you change the filter conditions, clear the cursor and start again from the first page. Mixing a cursor issued under one set of conditions with different conditions produces inconsistent results.

Understanding response fields

  • content: the answer body text.
  • sources: the list of sources the AI cited in the answer, with the URL, title, and domain of each. If you need the sentences actually quoted from a source, use include=quotes in Cited URL ranking.
  • fanouts: the list of search queries the AI issued internally to build the answer. See the fan-out description in Supported models.
  • markdownURL: URL of the raw answer as a markdown file.
fanouts and sources can be empty depending on the model and the answer. For when an empty value is expected behavior, see the empty-value rules in the Answer object.

Next steps

Analyzing with statistics

Summarize answers with source domain and URL rankings and fan-out aggregation