Research and fact-finding with ClawAI

What Research mode actually does

Research mode lets a request search the web, fetch a page, or fetch and extract structured content from one, before ClawAI produces an answer — so the answer can cite sources retrieved for that question rather than relying only on training data. It is a plan-gated feature with three depths: search only, search plus fetch, or search plus fetch and extract.

Metered separately from your model credit

Research access — web search, page fetches and extraction — is metered as its own usage, separate from the token allowance an ordinary chat message draws on. Your plan’s research allowance and its model token allowance are two different lines, not one shared pool, so running research does not draw down the credit a coding or writing task would use.

Picking a depth for the question at hand

A quick fact check usually needs only the search-only depth; a question that hinges on what a specific page actually says calls for search plus fetch; pulling structured data out of several pages at once is where search plus fetch and extract earns its cost. Matching the depth to the question keeps research usage proportionate rather than defaulting to the most expensive option every time.

Questions people ask

Does research use my model token credit?

No. Research access — web search, page fetches and extraction — is metered separately from the token allowance an ordinary chat message draws on. Confirm both allowances on the pricing page.

What is the difference between the three Research mode depths?

Search only returns results from a web search; search plus fetch also retrieves the page content; search plus fetch and extract additionally pulls structured content out of what it fetched.

Does the model I choose matter for research quality?

Yes — Research mode changes what sources a model can see, not how well it reads and reconciles them. See choosing a model for research with sources, linked below.