When a URL Analysis downloads a non-binary response (HTML, JSON, XML, plain text, …), this endpoint returns UTF-8 plain-text preview slices of that content.

The endpoint pages through the file: each request returns up to limit bytes starting at cursor, and reports a new cursor in the response so you can continue from where you left off.

{
  "data": "<preview text>",
  "meta": {
    "cursor": 65536
  }
}

Note that only the first page (cursor == 0) consumes intelligence download quota; subsequent paged calls do not.