Use AI Search when you need a focused answer with sources you can inspect. Use Research when you need to investigate a broader question, compare evidence, and assemble a structured report.
The distinction is the job you need done—not which mode has the more impressive name. A larger report is not automatically more useful than a concise answer.
Studio Global provides both modes. Here is how to choose between them using one illustrative business problem: evaluating software for a small team's customer onboarding process.
Compare the work, not just the output length
| Dimension | Search | Research |
|---|---|---|
| Starting question | A bounded factual question | A decision involving multiple questions |
| Example | What does the official setup guide require? | Which options fit our onboarding requirements? |
| Useful output | Answer, relevant evidence, and links | Comparison, synthesis, limitations, and next steps |
| Review focus | Does the source support this answer? | Are the sources and comparison criteria sufficient? |
These are usage recommendations, not measured speed, cost, or accuracy benchmarks. Model capabilities, access, and account limits can vary; check the current workspace controls.
Use Search to resolve a specific uncertainty
Suppose you already have a shortlist but need to know whether one product documents a particular export format. Open Search and ask:
Find current official documentation for [product]'s export formats.
Identify which formats are documented and any plan restrictions.
Link to the supporting pages and distinguish missing information
from an explicitly unsupported capability.
This keeps the task small. You can open the cited page and decide whether you have enough evidence to proceed.
If the answer says a feature was not found, treat that as a research gap. It is not proof of absence. Refine the query, consult the official documentation directly, or ask the vendor.
See the Search guide for citation review and follow-up questions.
Use Research to support a broader decision
Now suppose the team has not built a shortlist. It needs to compare setup, client experience, export options, and documented plan limitations. That is a wider investigation.
Research options for [team]'s customer onboarding process in [market].
Current process: [description]. Required outcomes: [requirements].
Compare [named options or selection criteria] using setup effort,
documented capabilities, plan restrictions, and evidence gaps.
Use official sources for product facts and independent context
where relevant. Separate facts, interpretation, and assumptions.
Return an executive summary, comparison table, and trial questions.
Set a decision horizon and request current evidence where freshness matters. A well-scoped report should expose tradeoffs, not just produce more paragraphs.
The Research guide explains how to frame the request and revise the result.
Move between the modes deliberately
You do not need to force the whole job into one mode. Start with Research to understand the options, then use Search to verify a decisive claim. Or begin with Search and move to Research when you discover that the question depends on several unresolved factors.
When moving between tasks, restate the decision and carry forward the evidence you need. Do not assume every earlier detail is automatically present in a new conversation.
Use a document when you need to edit the findings into a durable brief. A research report and an approved recommendation are different artifacts; the latter should reflect your team's priorities and review.
Know when neither mode is the best starting point
If the answer depends on an internal policy rather than public web information, start with an authorized file attachment or a relevant Infobase. Web research cannot establish an unpublished company rule.
If you only need to rewrite approved copy, give the AI the text and editing constraints. Searching the web may introduce unnecessary material into a task that already has its facts.
Whichever mode you choose, inspect important evidence, retain uncertainty, and avoid supplying sensitive information without authorization.
Start with Studio Global Search for your next focused question, then expand to Research when the decision calls for it.