20 Best Claude AI Prompts for In-Depth Analysis

Claude is useful for long documents, messy notes, research synthesis, and structured reasoning. It can still make mistakes, miss context, or overstate weak evidence. Treat every answer as analysis support, not final authority.

These prompts are designed to make Claude separate evidence from assumptions and produce outputs that are easier to review.

Anthropic’s prompt engineering docs recommend defining success criteria and empirical tests before prompt engineering. That is especially important for analysis work. A prompt is only good if the output helps you make a better decision, understand evidence more clearly, or identify what still needs checking.

1. Long Document Triage

Analyze this document for [purpose].

Document: [paste]
My goal: [goal]

Return:
- One-sentence summary
- Five key points
- What matters for my goal
- What is missing
- Claims that need verification
- Recommended next steps

2. Research Paper Review

Review this research paper.

Paper: [paste or summarize]
My background: [level]
Question I care about: [question]

Identify thesis, method, findings, limitations, practical implications, and claims that may be stronger than the evidence supports.

3. Evidence Synthesis

Synthesize these sources on [topic].

Sources: [paste summaries]

Separate:
- Areas of agreement
- Areas of disagreement
- Strong evidence
- Weak evidence
- Open questions
- Practical conclusion

4. Argument Deconstruction

Deconstruct this argument.

Text: [paste]

Identify conclusion, premises, assumptions, evidence quality, logical gaps, and the strongest counterargument.

5. Decision Analysis

Help me evaluate this decision.

Decision: [decision]
Options: [options]
Criteria: [criteria]
Constraints: [constraints]

Build a comparison table, list trade-offs, identify assumptions, and recommend what evidence would make the decision clearer.

6. Risk Review

Assess the risks in this plan.

Plan: [plan]
Context: [context]

For each risk, estimate likelihood, impact, early warning signs, mitigation, and owner. Flag risks I may be underestimating.

7. Market Signal Analysis

Analyze these market signals.

Signals: [data, notes, competitor moves]
Business context: [context]

Explain what the signals show, what they do not prove, possible explanations, and strategic implications.

8. Competitive Positioning

Compare our positioning with competitors.

Our product: [details]
Audience: [audience]
Competitors: [competitors]
Evidence: [proof]

Identify differentiation, weak claims, market gaps, and messaging we can support honestly.

9. Customer Feedback Themes

Analyze this customer feedback.

Feedback: [paste]
Segments: [optional]

Find themes, repeated pain points, emotional tone, feature requests, trust issues, and questions for follow-up research.

10. Survey Response Analysis

Analyze these survey responses.

Responses: [paste]
Goal: [goal]

Identify patterns, contradictions, segment differences, and what the survey cannot tell us.

11. Root Cause Analysis

Help diagnose this problem.

Problem: [problem]
Timeline: [timeline]
Data: [data]
Recent changes: [changes]

List possible causes, evidence for each, evidence against each, and the next test to run.

12. Process Bottleneck Analysis

Analyze this workflow for bottlenecks.

Workflow steps: [steps]
Symptoms: [symptoms]
Metrics: [metrics]

Identify likely constraints, verification methods, and low-risk improvements.

13. Stakeholder Perspective Map

Map stakeholder perspectives on [decision/change].

Stakeholders: [groups]
Context: [context]

For each group, identify priorities, concerns, likely objections, support conditions, and communication needs.

14. Scenario Planning

Create three scenarios for [topic].

Context: [context]
Key uncertainties: [uncertainties]

Build optimistic, base-case, and downside scenarios with triggers, implications, and actions to prepare.

15. Assumption Audit

Audit the assumptions behind this plan.

Plan: [plan]

List explicit assumptions, hidden assumptions, fragile assumptions, evidence needed, and what changes if each assumption fails.

16. Text Sentiment and Intent

Analyze this message for tone, intent, and possible subtext.

Message: [paste]
Relationship/context: [context]

Separate what is directly stated from reasonable inference. Avoid over-reading.

17. Strategy Memo

Turn this analysis into a strategy memo.

Inputs: [paste]
Audience: [audience]
Decision needed: [decision]

Write an executive summary, evidence, options, risks, recommendation, and open questions.

18. Red Team Review

Red-team this proposal.

Proposal: [paste]

Find the strongest objections, failure modes, weak evidence, stakeholder pushback, and revisions that would make it stronger.

19. Retrospective Analysis

Analyze this completed project.

Goal: [goal]
Actual outcome: [outcome]
Timeline: [timeline]
What changed: [changes]

Identify what worked, what failed, what we learned, and what to do differently next time.

20. Verification Plan

Create a verification plan for this analysis.

Analysis: [paste]

List facts to verify, sources to check, experts to consult, data needed, and conclusions that should remain tentative.

Best Practices

Use real source material. Ask Claude to label assumptions. Request a verification plan. For high-stakes decisions, compare Claude’s output with human experts, primary sources, and original data.

How to Use These Prompts Well

Claude tends to work best when the task is specific and the source material is clear. Instead of asking “analyze this,” tell it what decision, risk, audience, or question matters. If you are analyzing a report for a CEO, the output should be different from an academic summary or a product team retrospective.

Add:

The goal of the analysis.
The audience.
The decision to support.
The source material.
The output format.
What uncertainty should be flagged.

Prompt 21: Source-Limited Analysis

Analyze only the source material below.
If the source does not support a claim, say "not supported by the source."

Source:
[paste]

Question:
[question]

Use this when accuracy matters more than creativity. It is especially helpful for policies, contracts, research excerpts, meeting notes, and internal documents.

Prompt 22: Executive Briefing

Turn this material into an executive briefing.

Inputs: [paste]
Audience: [audience]
Decision needed: [decision]

Include:
1. Executive summary.
2. Key evidence.
3. Options.
4. Risks.
5. Recommendation.
6. Open questions.

This prompt is useful when the output needs to be short, decision-ready, and careful.

Prompt 23: Evidence Strength Rating

Rate the evidence behind each major claim.

Claims: [paste]
Sources: [paste]

For each claim, label evidence as strong, moderate, weak, or unsupported.
Explain the label briefly.

Evidence labels help prevent confident summaries from hiding weak support.

Prompt 24: Assumption Stress Test

Stress-test these assumptions.

Assumptions: [list]
Plan: [plan]

For each assumption:
1. What would make it false?
2. How would we notice early?
3. What is the impact if it fails?
4. What backup plan should we prepare?

This is useful for strategy, hiring plans, product roadmaps, budget decisions, and go-to-market plans.

Review this analysis for bias and blind spots.

Analysis: [paste]
Context: [context]

Look for:
1. Confirmation bias.
2. Missing stakeholders.
3. Overreliance on weak data.
4. Ignored alternatives.
5. Overconfident language.
6. What a skeptical expert would challenge.

This prompt is a useful second pass before sharing analysis with a team.

When to Use Claude vs Other Tools

Use Claude when you need careful reading, structured synthesis, long-form critique, or nuanced explanation. Use spreadsheets, notebooks, databases, or statistical tools when you need exact calculations. Use current web research when the facts may have changed. Use human experts when the decision is high stakes.

Claude can help think through material. It should not become the only source of truth.

Review Checklist

Before acting on Claude’s analysis:

Did it cite or reference the source material correctly?
Did it separate facts from assumptions?
Did it overstate the evidence?
Did it miss obvious stakeholders?
Did it identify uncertainty?
Did a human review the original material?
Are current facts verified?
Are high-stakes conclusions checked by experts?

Analysis Workflow by Use Case

The best prompt depends on the kind of analysis you are doing. A research task needs source discipline. A strategy task needs trade-offs. A customer-feedback task needs pattern detection without losing emotional nuance. A risk task needs early warning signs and owners.

For research, start with document triage, then evidence synthesis, then a verification plan. Ask Claude to identify what the sources actually support and what is still inference. This reduces the risk of turning a polished summary into a false conclusion.

For strategy, begin with decision analysis, competitive positioning, and scenario planning. Claude can help expose hidden assumptions, but the team still needs to supply business context, constraints, and actual market evidence.

For customer feedback, use theme analysis first, then stakeholder perspective mapping. Ask Claude to keep quotes or examples attached to each theme when possible. That makes the analysis easier to audit and prevents the model from smoothing away sharp customer language.

For risk reviews, use risk review, assumption audit, and red-team prompts together. The output should not only list risks. It should explain how each risk would appear early, who owns it, and what mitigation is realistic.

For document review, use source-limited analysis. Tell Claude not to use outside knowledge unless you explicitly ask for it. This is useful when reviewing internal policies, notes, proposals, contracts, or technical documentation where the source text is the authority.

Claude Prompt Template You Can Reuse

Use this template when you do not know where to start:

You are helping me analyze [material].

Goal:
[What decision, question, or understanding I need]

Audience:
[Who will use the output]

Source material:
[Paste or summarize the material]

Constraints:
[Time, budget, policy, risk, required format]

Please return:
1. Short summary.
2. Main findings.
3. Evidence for each finding.
4. Assumptions.
5. Uncertainties.
6. Counterarguments.
7. Recommended next steps.
8. Facts I should verify manually.

This structure works because it gives Claude a job, a reader, source material, boundaries, and an output format. It also asks for uncertainty directly, which is one of the simplest ways to get more responsible analysis.

Common Mistakes

The biggest mistake is asking Claude for “deep analysis” without giving it enough material. A model cannot analyze evidence it has not seen. If you want better output, give it the document, notes, data sample, interview excerpts, or decision criteria.

The second mistake is accepting the first answer. A better workflow is to ask for analysis, then ask Claude to critique its own answer, then ask what evidence would change the conclusion. This second pass often reveals weak assumptions.

The third mistake is mixing creative brainstorming with factual analysis. If you want ideas, invite speculation. If you want evidence, restrict the model to the source. Those are different jobs.

The fourth mistake is treating confidence as accuracy. Claude may write fluently even when a point needs checking. Strong formatting is not proof.

Final Recommendation

Use Claude as an analysis partner, not an oracle. It is excellent at organizing messy information, finding patterns, drafting structured memos, and showing you where to look next. It is weaker when the task requires real-time facts, exact calculations, proprietary context it has not been given, or professional judgment in high-stakes domains.

The strongest workflow is simple: provide evidence, ask for structure, require uncertainty, verify important facts, and let humans make the decision.

References

FAQ

Can Claude analyze long documents accurately?

It can help summarize and organize long documents, but important findings should be checked against the original text.

Should I paste confidential documents?

Only if your tool, plan, and organizational policies permit it. For sensitive work, use approved enterprise settings and remove unnecessary personal or confidential information.

How do I avoid shallow analysis?

Ask for evidence, assumptions, counterarguments, and what would change the conclusion.

Conclusion

Claude is strongest when you give it a clear analytical job and enough evidence to work from. Use it to structure thinking, find patterns, and create reviewable drafts. Keep human verification in charge.

20 Best Claude AI Prompts for In-Depth Analysis