Claude Research is intended for deeper, multi-source tasks and may depend on a paid plan or organization settings.
1. Write the research decision first
State what you are trying to decide or produce, not just the topic.
2. Set boundaries
Include timeframe, geography, source preferences, excluded sources and any internal documents that should be considered.
3. Start Research when available
Use the Research feature for a multi-step investigation; keep normal chat/search for narrower questions.
4. Challenge the synthesis
Ask which claims are directly supported, which are inferred, and where reputable sources disagree.
5. Read the critical sources
Open the sources behind the conclusion and check whether Claude preserved the original context, dates and qualifiers.
Research [question] for [decision/output]. Scope: [dates, geography]. Prefer [source types]. Exclude [weak sources]. For each major conclusion, mark it as directly supported, inferred, or disputed. Include a short “what would change this conclusion?” section.
Asking only for “everything about X”; not defining recency; accepting a single synthesized narrative when the evidence is mixed; failing to distinguish company claims from independent evidence.
Finish check
The finished research should show an evidence trail strong enough that another reader can inspect the same sources and understand how the conclusion was formed.