Reasoning-Before-Conclusions & Verifiable Sourcing

Ask for Reasoning Before the Answer, When It's Warranted

For analysis and multi-step reasoning tasks, asking Claude to work through its reasoning before giving a final answer generally improves quality -- it gives the model room to weigh evidence and check its own logic before committing to a conclusion. For simple lookups or short drafts, that overhead isn't needed and just adds length without improving the result.

Research Needs Verifiable Sourcing

Research-style asks carry a specific risk: Claude can produce confident-sounding output that is not actually verified. A research prompt should therefore explicitly ask for structured findings with sources or citations you can check yourself, and should state the scope and recency required (e.g., "only sources from the last two years"). Confidence in the phrasing of an answer is not evidence that the underlying claim is correct.

Why This Belongs in Task-Type Adaptation

These two levers -- reasoning-before-conclusions and verifiable sourcing -- are the specific mechanics behind the general "analysis" and "research" rows of the task-type table: they explain *why* those task types need that particular emphasis, not just that they do.

Common exam traps

• Forgetting that research-style asks need verifiable sourcing, since Claude can produce confident but unverified claims. Treating a fluent, well-organized research answer as automatically trustworthy skips the verification step the task type actually requires.
• Reflexively asking for step-by-step reasoning on every task, including simple lookups or short drafts where it adds no benefit and only adds length.