How web research works: scope, review, and results

Web research finds and reads public pages to help answer a Task. Ask a focused question and request the source details your application needs to review the answer. The result can include supporting sources and unresolved gaps; those sources still need to be checked against the claim.

From a research question to a source-supported answerA focused question guides research into public pages. Selected sources provide evidence and reveal gaps. The Task Result can include the answer, source links, and unresolved questions when the requested response model includes those fields.

What defines the research scope

A question’s subject, date, and source requirements define which evidence is relevant. The request guide shows how to set that scope.

Run a bounded source review

Web research uses search to discover candidate pages and reads selected sources for evidence. A search extract alone does not establish a claim. The review is bounded by the Task’s available time and source access; it is not an exhaustive survey of the Web.

What the Task Result contains

Your terminal SDK result follows the response model that your application requested. It can contain the answer, source links, supporting excerpts, and unresolved gaps when the model includes those fields. The request guide gives an example.

Citations make an answer easier to inspect; they do not establish that it is correct. Unresolved gaps remain missing evidence. Verify a material claim describes the review needed before relying on it.

Understand the data boundary

Details from your request can become public search queries. Include only the detail needed to find the evidence, and do not include secrets.

Data protection covers query and page-request paths. Page access covers supported page behavior. Troubleshoot Web research covers missing or unreadable sources.