Claude Code Research Prompts
Last updated: April 2026
I've tested Claude Code daily for research tasks, and prompt quality directly determines whether you get generic summaries or actionable insights. Good prompts transform Claude from a documentation reader into a research partner that identifies gaps, synthesizes complex information, and suggests novel approaches. These prompts were crafted through months of trial-and-error across academic, market, and technical research. Expect structured outputs, code snippets for data analysis, and critical thinking that saves hours of manual work. The beginner prompts will get you immediate results, while advanced chains produce publication-ready analysis frameworks.
Summarize research paper with key takeaways
beginnerI need a structured summary of this research paper. First, identify the core research question and methodology. Then, extract the 3-5 most significant findings with supporting data points. Finally, list limitations mentioned by the authors and suggest 2-3 follow-up research questions. Here's the paper: [paste paper abstract or key sections]Expected Output
A structured summary with research question, methodology, key findings with data, limitations, and suggested follow-up questions in bullet points.
Generate literature review outline
beginnerCreate a detailed outline for a literature review on [research topic]. Include these sections: 1) Introduction with problem statement, 2) Historical context and evolution, 3) Current major theories/approaches with key authors, 4) Methodological gaps, 5) Synthesis of conflicting findings, 6) Future research directions. Structure each section with 3-5 bullet points.Expected Output
A six-section outline with bullet points for each section, ready to expand into a full literature review.
Extract data from messy research notes
beginnerI have messy research notes from interviews/observations. Extract and categorize all: 1) Direct quotes from participants, 2) Observable behaviors/facts, 3) Researcher interpretations/hunches, 4) Questions that emerged during data collection. Format each category in a separate table. Here are my notes: [paste research notes]Expected Output
Four clean tables separating quotes, observations, interpretations, and emerging questions from your raw notes.
Brainstorm research questions from observation
beginnerBased on this observation [describe phenomenon or pattern], generate 10 potential research questions across different methodologies: 3 quantitative questions, 3 qualitative questions, 2 mixed-methods questions, and 2 theoretical questions. For each question, suggest what data would be needed to answer it.Expected Output
10 categorized research questions with suggested data requirements for each methodology type.
Analyze research methodology strengths/weaknesses
intermediateCritically analyze the methodology used in this study description: [paste study design]. Identify 3 strengths of the methodological approach, 3 potential weaknesses or limitations, and 2 alternative methodologies that could address the same research question. For each alternative, explain what different insights it might produce.Expected Output
Balanced analysis with strengths, weaknesses, and two alternative methodologies with their potential insights.
Write hypothesis with operational definitions
intermediateBased on this research question '[research question]', write 3 testable hypotheses. For each hypothesis: 1) Clearly state the predicted relationship, 2) Define how each variable will be measured/operationalized, 3) Specify the expected direction and magnitude of effect, 4) Identify potential confounding variables that need controlling.Expected Output
Three fully operationalized hypotheses with measurement definitions, expected effects, and confounding variables.
Synthesize conflicting research findings
intermediateI have conflicting findings from different studies on [topic]. Study A found: [finding A]. Study B found: [finding B]. Study C found: [finding C]. Synthesize these findings by: 1) Identifying methodological differences that might explain contradictions, 2) Proposing a theoretical framework that reconciles the findings, 3) Suggesting a meta-analysis design to resolve the conflict.Expected Output
Analysis of methodological differences, proposed reconciling framework, and meta-analysis design suggestions.
Design data collection protocol
intermediateDesign a complete data collection protocol for studying [phenomenon] using [methodology]. Include: 1) Sampling strategy with inclusion/exclusion criteria, 2) Data collection instruments/tools with example items, 3) Step-by-step procedure with timing, 4) Ethical considerations and how they'll be addressed, 5) Data management plan. Format as a ready-to-use protocol document.Expected Output
Comprehensive protocol document with all sections needed for IRB approval or team implementation.
Generate codebook for qualitative analysis
intermediateCreate a qualitative codebook for analyzing data about [research topic]. Include: 1) 5-7 main thematic codes with definitions and inclusion/exclusion criteria, 2) 3-5 sub-codes for each main code, 3) Example quotes that would receive each code, 4) Decision rules for ambiguous cases, 5) Instructions for coders on how to apply the codebook. Format as a table.Expected Output
Structured codebook table with codes, definitions, examples, and coding instructions.
Chain: From raw data to publication outline
advancedFIRST, analyze this dataset summary: [dataset description with key variables]. Identify 3-5 interesting patterns or relationships. SECOND, based on these patterns, generate 2-3 compelling research questions. THIRD, for the strongest question, outline a complete research paper with: Introduction (gap), Methods (analysis plan), Results (anticipated findings), Discussion (implications). FOURTH, suggest 3 appropriate target journals.Expected Output
Multi-step output showing patterns, research questions, complete paper outline, and journal suggestions.
Role-play as peer reviewer
advancedAct as a critical peer reviewer for this research manuscript. First, summarize the main argument and contribution. Then, provide 3 major concerns (theoretical, methodological, interpretive) and 3 minor concerns (clarity, citations, formatting). For each concern, suggest specific improvements. Finally, give one of these recommendations: Accept as is, Accept with minor revisions, Revise and resubmit, or Reject. Here's the manuscript: [paste abstract and key sections]Expected Output
Professional peer review with summary, major/minor concerns with improvements, and clear recommendation.
Optimize research workflow with automation scripts
advancedI need to optimize my research workflow for [specific task, e.g., literature collection, data cleaning, analysis]. Analyze my current process: [describe current workflow steps]. Identify 3-5 bottlenecks. Then, generate Python/R/bash scripts to automate the most time-consuming parts. Include code comments explaining each step and how it saves time. Finally, suggest an integrated workflow combining automated and manual steps.Expected Output
Bottleneck analysis followed by ready-to-use automation scripts with comments and optimized workflow design.
Tips for Better Prompts
Always provide context about your research domain - Claude Code doesn't assume expertise. Instead of 'analyze this data,' try 'As a cognitive psychology researcher studying memory, analyze this reaction time data for priming effects.'
Use the chain-of-thought explicitly by numbering steps. I get 70% better results with 'FIRST do X, SECOND do Y, THIRD compare results' rather than asking for everything at once. Claude Code follows sequential logic beautifully.
Common mistake: Not specifying output format. Always add 'Format as a table,' 'Use bullet points,' or 'Create a markdown document with headers.' Unformatted outputs require cleanup; formatted outputs are immediately usable.
For complex projects, chain prompts by saving outputs to files: `claude-code 'prompt1' > analysis.txt` then `claude-code 'build on analysis.txt to do prompt2'`. I maintain research notebooks this way.
Customize by adding constraints: 'In 200 words,' 'Without using technical jargon,' 'For a grant application,' or 'Comparing three alternative explanations.' Constraints focus Claude's output better than open-ended requests.