20 Best Claude AI Prompts for In-Depth Analysis

The fastest way to get better analysis from Claude? Give it a clear job, real source material, and explicit output constraints. These 20 templates were tested against Claude Opus 4.7 and Claude Sonnet 4.6 and produce accurate, structured outputs.

Claude leads on analysis because of superior format adherence and self-correction. The 200K token context handles ~500 pages paste full documents, not summaries.

Quick comparison: Claude vs. GPT-5.1 vs. Gemini 3 for analysis

Capability	Claude Opus 4.7	GPT-5.1	Gemini 3
Instruction following	Excellent	Very Good	Good
Format adherence	Excellent	Very Good	Good
Long context	200K tokens	128K	1M tokens
Self-correction	Strong	Strong	Moderate
Evidence verification	Excellent	Very Good	Good

“Claude performs best when you give it clear success criteria, structured inputs, and explicit output constraints.” Anthropic Prompt Engineering Guide, 2026

How to Use This Guide

Each prompt follows the four-block pattern: INSTRUCTIONS, CONTEXT, TASK, and OUTPUT FORMAT.

Key rule: Paste full documents, not summaries. Claude performs measurably worse from summaries.

The 20 Prompts

1. Long Document Triage

## INSTRUCTIONS
You are a senior research analyst. Review the document below and provide structured findings. If the source does not support a claim, say "not supported by the source."

## CONTEXT
My goal: [what decision or understanding I need]
Document: [paste full document]

## TASK
Return: 1) One-sentence summary, 2) Five key findings, 3) What matters for my goal, 4) Gaps or missing info, 5) Claims needing verification, 6) Next steps

## OUTPUT FORMAT
Numbered sections. Mark uncertain claims [UNCERTAIN].

2. Research Paper Review

## INSTRUCTIONS
You are a critical reviewer. Evaluate thesis, methodology, findings, and limitations. Flag claims stronger than the evidence.

## CONTEXT
My background: [novice/intermediate/expert]
My question: [specific question]
Paper: [paste paper]

## TASK
Return: 1) Thesis, 2) Methodology assessment, 3) Key findings, 4) Limitations, 5) Relevance to my question, 6) Overstated claims

## OUTPUT FORMAT
Use headings. Rate evidence: Strong / Moderate / Weak / Unsupported.

3. Evidence Synthesis

## INSTRUCTIONS
You are a synthesis analyst. Compare sources. Separate factual agreement from interpretive disagreement.

## CONTEXT
Topic: [topic]
Sources: [paste or summarize]

## TASK
Return: 1) Areas of agreement, 2) Areas of disagreement, 3) Strongest evidence, 4) Weakest claims, 5) Open questions, 6) Practical conclusion

## OUTPUT FORMAT
Bullet lists. Mark uncertain items [INFERRED] or [UNVERIFIED].

4. Argument Deconstruction

## INSTRUCTIONS
You are a logical analysis specialist. Break down the argument into conclusion, premises, assumptions, and evidence quality. Find the strongest counterargument.

## CONTEXT
Argument: [paste]

## TASK
Return: 1) Conclusion, 2) Premises, 3) Explicit assumptions, 4) Hidden assumptions, 5) Evidence quality, 6) Logical gaps, 7) Strongest objection

## OUTPUT FORMAT
Structured breakdown with sections.

5. Decision Analysis

## INSTRUCTIONS
You are a decision analyst. Evaluate options against criteria. Surface trade-offs, assumptions, and what evidence would clarify the decision.

## CONTEXT
Decision: [decision]
Options: [options]
Criteria: [criteria ranked]
Constraints: [budget, time, policy]

## TASK
Return: 1) Comparison table, 2) Trade-offs per option, 3) Key assumptions, 4) Evidence needed, 5) Recommended option with caveats, 6) Risks

## OUTPUT FORMAT
Markdown table. Bullets for trade-offs.

6. Risk Review

## INSTRUCTIONS
You are a risk management specialist. Assess risks. Flag any that may be underestimated.

## CONTEXT
Plan: [plan]
Context: [organizational context]

## TASK
For each risk: 1) Likelihood, 2) Impact, 3) Warning signs, 4) Mitigation, 5) Owner, 6) Underestimated flag

## OUTPUT FORMAT
Bolded risk names. Summary table: Likelihood | Impact | Status.

7. Market Signal Analysis

## INSTRUCTIONS
You are a market analyst. Interpret signals and explain what they do not prove. Avoid overclaiming.

## CONTEXT
Business context: [company situation]
Signals: [data, competitor moves, customer behavior]

## TASK
Return: 1) What signals suggest, 2) What they do not prove, 3) Alternative explanations, 4) Strategic implications, 5) Data to strengthen confidence

## OUTPUT FORMAT
Sections. Mark speculative [SPECULATIVE].

8. Competitive Positioning

## INSTRUCTIONS
You are a competitive strategy analyst. Compare positioning honestly. Identify real differentiation and messaging needing stronger evidence.

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

## TASK
Return: 1) Genuine differentiation, 2) Overclaimed positioning, 3) Market gaps, 4) Defensible messaging, 5) Messaging needing evidence

## OUTPUT FORMAT
Bullet lists. Flag unsupported claims [UNSUPPORTED].

9. Customer Feedback Themes

## INSTRUCTIONS
You are a customer insights analyst. Find themes, pain points, emotional signals, and feature requests. Preserve quotes illustrating each theme.

## CONTEXT
Segments: [optional]
Feedback: [paste]

## TASK
Return: 1) Major themes, 2) Pain points with quotes, 3) Positive signals, 4) Feature requests, 5) Trust issues, 6) Follow-up questions

## OUTPUT FORMAT
Organize by theme. Attach [QUOTE] tags for direct language.

10. Survey Response Analysis

## INSTRUCTIONS
You are a survey analyst. Find patterns, contradictions, and segment differences. State what the survey cannot tell us.

## CONTEXT
Goal: [what survey was designed to learn]
Responses: [paste]

## TASK
Return: 1) Key patterns, 2) Contradictions, 3) Segment differences, 4) What survey answers definitively, 5) What it cannot answer, 6) Follow-up questions

## OUTPUT FORMAT
Markdown tables. Mark [SURVEY-SUPPORTED] or [INFERRED].

11. Root Cause Analysis

## INSTRUCTIONS
You are a diagnostic analyst. List causes and evaluate evidence for and against each. Recommend the next diagnostic test.

## CONTEXT
Problem: [description]
Timeline: [events]
Data: [available data]
Recent changes: [changes]

## TASK
Return: 1) Possible causes ranked by likelihood, 2) Evidence for each, 3) Evidence against each, 4) Next test for each, 5) Most likely cause

## OUTPUT FORMAT
Bolded cause names with evidence bullets.

12. Process Bottleneck Analysis

## INSTRUCTIONS
You are a process improvement analyst. Find bottlenecks, suggest verification methods, and propose low-risk improvements.

## CONTEXT
Workflow steps: [steps]
Symptoms: [observed problems]
Metrics: [metrics]

## TASK
Return: 1) Likely bottlenecks ranked by impact, 2) Verification method, 3) Low-risk improvement, 4) Expected impact

## OUTPUT FORMAT
Numbered bottleneck list. Summary table.

13. Stakeholder Perspective Map

## INSTRUCTIONS
You are a stakeholder strategy analyst. Map perspectives for each group: priorities, concerns, objections, communication needs.

## CONTEXT
Stakeholders: [groups]
Decision: [context]
Context: [culture, history]

## TASK
For each group: 1) Priorities, 2) Concerns, 3) Likely objections, 4) Support conditions, 5) Preferred communication

## OUTPUT FORMAT
Bolded group headings. Bullets per attribute.

14. Scenario Planning

## INSTRUCTIONS
You are a scenario planner. Build three scenarios (optimistic, base-case, downside) with triggers, implications, and actions.

## CONTEXT
Topic: [topic]
Key uncertainties: [uncertainties]
Time horizon: [timeframe]

## TASK
Build three scenarios with: 1) Trigger, 2) Implications, 3) Preparation actions. Identify which uncertainty each depends on.

## OUTPUT FORMAT
Clear headings per scenario. Bullets for triggers, implications, actions.

15. Assumption Audit

## INSTRUCTIONS
You are a strategy auditor. Audit the plan for explicit and hidden assumptions. Identify what would make each false.

## CONTEXT
Plan: [plan description]

## TASK
Return: 1) Explicit assumptions, 2) Hidden assumptions, 3) Fragile assumptions, 4) Evidence needed, 5) Impact if each fails

## OUTPUT FORMAT
Two columns: assumption and validation. Mark fragile [FRAGILE].

16. Text Sentiment and Intent

## INSTRUCTIONS
You are a communication analyst. Analyze tone, intent, and subtext. Separate stated content from inference. Avoid over-reading.

## CONTEXT
Message: [paste]
Relationship: [context]
Setting: [context]

## TASK
Return: 1) Direct content, 2) Tone, 3) Likely intent, 4) Possible subtext, 5) Recommended response

## OUTPUT FORMAT
Bolded headings. Distinguish direct from inference.

17. Strategy Memo

## INSTRUCTIONS
You are a strategic communications specialist. Convert analysis into a decision-ready executive memo.

## CONTEXT
Analysis: [paste]
Audience: [who reads this]
Decision: [what this supports]

## TASK
Write memo: 1) Executive summary (3 sentences), 2) Key evidence, 3) Options, 4) Risks, 5) Recommendation with confidence, 6) Open questions

## OUTPUT FORMAT
Memo format with bolded section headers.

18. Red Team Review

## INSTRUCTIONS
You are a red team analyst. Find the strongest objections, failure modes, and weak evidence. Suggest revisions to strengthen.

## CONTEXT
Proposal: [paste]

## TASK
Return: 1) Strongest objection, 2) Failure modes, 3) Weak evidence, 4) Likely stakeholder pushback, 5) Specific revisions

## OUTPUT FORMAT
Bolded headings. Be direct.

19. Retrospective Analysis

## INSTRUCTIONS
You are a learning analyst. Identify what worked, what failed, what was learned, and what to do differently.

## CONTEXT
Goal: [original goal]
Outcome: [what happened]
Timeline: [project timeline]
Changes: [scope, external factors]

## TASK
Return: 1) What worked (with reasons), 2) What did not work (with reasons), 3) Key lessons, 4) Changes for next time, 5) What to do differently today

## OUTPUT FORMAT
Three sections: What Worked, What Did Not Work, What to Change.

20. Verification Plan

## INSTRUCTIONS
You are a research verification specialist. Create a verification plan prioritizing claims by importance.

## CONTEXT
Analysis to verify: [paste]

## TASK
Return: 1) Facts to verify (prioritized), 2) Sources to check, 3) Experts to consult, 4) Data needed, 5) Tentative conclusions, 6) Verification order

## OUTPUT FORMAT
Numbered list with bolded facts. Mark tentative [TENTATIVE].

Best Practices

Three factors most improve Claude’s analysis quality:

State goal and constraints first. Define what “done” looks like. “Summarize in 3 bullets” outperforms vague instructions.
Paste source material directly. Full documents, not summaries. Claude’s 200K context handles 500 pages.
Force structure in output. Use markdown tables or numbered sections. Prevents skipping sections.

The 80/20 checklist: Define success criteria, paste source material, request uncertainty flags, add “if unsure, say so”, specify output format.

Prompt Comparison Table

Task Type	Best Model	Effort	Key Element
In-depth analysis	Opus 4.7	High/Xhigh	Source + verification
Summarization	Sonnet 4.6	Medium	Long context placement
Evidence synthesis	Opus 4.7	High	Multiple source comparison
Strategic memos	Sonnet 4.6	High	Structured output schema
Red team / critique	Opus 4.7	Xhigh	Explicit objection request
Root cause diagnosis	Sonnet 4.6	High	Evidence ranking format

FAQ

Can Claude analyze long documents accurately?

Claude processes up to 200K tokens in one pass. Paste full documents, not summaries.

Should I paste confidential documents?

Only if tool configuration, policies, and data handling agreements permit it. Use Anthropic’s enterprise configurations for sensitive work.

How do I avoid shallow analysis?

Ask Claude to label evidence strength. Request assumptions and counterarguments. Add “If unsure, say so explicitly.”

What effort setting for analysis tasks?

Use high or xhigh for intelligence-sensitive work. Medium for routine summarization.

When Opus 4.7 vs. Sonnet 4.6?

Use Opus 4.7 for complex analysis and high-stakes decisions. Use Sonnet 4.6 for routine tasks where cost matters more.

Can Claude replace human experts?

No. Claude excels at organizing information, finding patterns, drafting structured memos. It is weaker on real-time facts, exact calculations, proprietary context, or professional judgment in high-stakes domains.

Sources

Anthropic Docs: Prompt Engineering Overview (accessed May 2026)
Anthropic Docs: Claude Prompting Best Practices (accessed May 2026)
Anthropic Docs: Claude Models Overview (accessed May 2026)
Prompt Builder: Claude Prompt Engineering Best Practices 2026 (December 13, 2026)
AI Prompt Library: 30 Claude Data Analysis Prompts (updated May 2026)

AI Unpacker provides practical, tested guidance on using AI tools effectively. All prompts tested against Claude Opus 4.7 and Claude Sonnet 4.6.

Key Takeaways

Summarize with AI