Skip to main content
The model value is GOOGLE_OVERVIEW. Any behavior not listed here follows the Supported models table as-is.

Where answers are collected

Google Overview answers are collected from the “AI Overview” shown at the top of the Google search results page.
An AI Overview block shown at the top of Google search results

The Google search results screen used for collection. The prompt is entered as the search query, and the body and sources of the 'AI Overview' block at the top are collected.

Collection conditions are as follows.
  • Collection happens without signing in. An account’s search history and personalization settings do not affect the results.
  • The prompt is entered verbatim as the search query. It is not rewritten or shortened to keywords.
  • Regular search results in the “All” tab are used. The “AI Mode” tab is collected by a separate model, Google AI Mode.
  • Region and location are applied by connecting from a network that matches the run request’s region. Whether Google shows an AI Overview, and what it contains, varies greatly by connection region.
Screen elements map to answer object fields as follows.
Some questions never get an AI Overview. Google does not show an AI Overview for every search query. For such prompts there is no answer to collect, so the run ends as FAILED. This is the expected outcome, not a collection error. No credits are deducted. A temporary collection failure also ends as FAILED, and the response alone does not tell the two apart. Run the same prompt once more; if it ends as FAILED again, treat it as a question that gets no AI Overview.

Model-specific behavior

Fan-out is not provided. fanouts is always an empty array, while the answer body and cited sources are collected normally.