Best Claude Prompts: 50 Examples That Actually Work (2026)

Bottom Line Up Front

Claude responds best to prompts that are direct, structured, and context-rich. Unlike models trained primarily on reinforcement learning from human feedback (RLHF) alone, Claude uses Constitutional AI (CAI), which means it self-corrects against a set of principles during training. The practical result: Claude handles nuanced instructions, multi-step reasoning, and long-form output more reliably than most alternatives — but only when you prompt it correctly. According to Grokipedia’s Claude entry, Anthropic’s Claude family has processed over 10 billion API requests since its public launch, and prompt quality is the single biggest factor in output quality. A 2026 McKinsey survey found that 67% of professionals using AI assistants say “knowing how to prompt well” is now a core job skill. The 50 prompts below are tested, specific, and ready to paste into Claude right now.

Key Takeaways

• 50 copy-paste-ready prompts across six professional categories, each tested on Claude 3.5 Sonnet and Claude 3 Opus
• Claude-specific optimization — these prompts leverage Claude’s 200K context window, Constitutional AI training, and instruction-following strengths
• Each prompt includes four elements: the exact text, what it does, why it works with Claude specifically, and a customization tip
• Prompts are organized by use case: writing (10), business (10), research (10), coding (10), education (5), creative (5)
• Prompt engineering matters more than model selection — a well-crafted prompt on Claude 3.5 Sonnet regularly outperforms a vague prompt on more expensive models

Why Claude Prompts Are Different

Before diving into the 50 prompts, it helps to understand why Claude responds differently than ChatGPT, Gemini, and other models. Claude’s training uses Constitutional AI, a method developed by Anthropic where the model critiques and revises its own outputs against a set of written principles. This creates three practical differences for prompt engineering:

1. Claude follows complex, multi-part instructions more faithfully. You can give Claude a 10-step prompt and expect it to address every step. According to Anthropic’s Constitutional AI paper (arXiv:2212.08073), this instruction-following capability is a direct result of the self-critique training loop.
2. Claude handles longer contexts better. With a 200K-token context window, you can paste entire documents, codebases, or datasets directly into your prompt. This changes how you structure prompts — instead of summarizing context, you can provide the full source material.
3. Claude is more receptive to role-based and persona prompts. When you tell Claude to “act as a senior data analyst,” it commits to that framing more consistently throughout long outputs than models without CAI training.

If you are new to working with Claude, start with our complete guide to using Claude AI for setup basics, then return here for the prompt library. For foundational prompt engineering concepts that apply across all models, see our guide to writing AI prompts.

---

Writing Prompts (1–10)

Claude excels at writing tasks because its Constitutional AI training produces outputs that are clear, well-organized, and tonally consistent. These 10 prompts cover the most common professional writing needs.

1. Blog Post Writer

Write a 1,500-word blog post about [TOPIC]. Target audience: [AUDIENCE]. Tone: conversational but authoritative. Structure it with an introduction that hooks the reader with a specific statistic or surprising fact, 4-5 H2 subheadings that each cover one distinct subtopic, bullet points for scannable takeaways, and a conclusion with a clear call to action. Include at least 3 specific data points or examples. Do not use filler phrases like "in today's fast-paced world" or "it's no secret that."

What it does: Generates a complete, SEO-friendly blog post with proper structure, real substance, and a specific word count target.

Why it works with Claude: Claude’s instruction-following means it will actually hit all the structural requirements (subheadings, bullet points, CTA) in a single pass. The explicit ban on filler phrases leverages Claude’s ability to follow negative constraints — something it handles better than most models due to CAI training.

Customization tip: Replace [TOPIC] and [AUDIENCE] with specifics. Add “Write in first person” or “Write in third person” to control voice. For longer posts, increase the word count and add more H2 sections proportionally.

2. Email Drafter

Draft a professional email for the following situation: [DESCRIBE THE SITUATION, THE RECIPIENT, AND YOUR GOAL]. The email should be no longer than 200 words. Use a [formal/friendly/direct] tone. Include a clear subject line, one specific ask or call to action, and close with a concrete next step (not a vague "let me know"). If there are any sensitive or potentially awkward aspects to navigate, handle them diplomatically.

What it does: Creates a concise, purposeful email that gets straight to the point with a specific call to action.

Why it works with Claude: Claude’s diplomatic and nuanced tone is one of its strongest differentiators. The instruction to “handle sensitive aspects diplomatically” plays to Claude’s CAI training, which specifically optimizes for helpfulness while avoiding harm. Claude produces emails that sound human and tactful — not robotic.

Customization tip: Paste the previous email thread into the prompt for reply drafting. Claude’s 200K context window means you can include an entire email chain without truncation.

3. Social Media Caption

Write 5 social media captions for [PLATFORM] about [TOPIC/PRODUCT]. Each caption should: be under [CHARACTER LIMIT] characters, include a hook in the first line that stops scrolling, use [BRAND VOICE — e.g., witty, inspiring, educational], end with a clear CTA, and include 3-5 relevant hashtags. Format: number each caption and put the hashtags on a separate line below each one.

What it does: Generates multiple caption options with platform-appropriate formatting, hooks, and engagement elements.

Why it works with Claude: By asking for 5 variations, you leverage Claude’s ability to maintain variety while following the same constraints. Claude avoids repeating the same hook structure across variations — a common failure mode in other models.

Customization tip: Specify the platform character limits (X/Twitter: 280, LinkedIn: 3,000, Instagram: 2,200) and Claude will respect them precisely.

4. Product Description

Write a product description for [PRODUCT NAME]. Details: [LIST KEY FEATURES, SPECS, PRICE, TARGET CUSTOMER]. The description should be 150-200 words, lead with the primary benefit (not a feature), include 3 specific features that map to customer pain points, use sensory or emotional language where appropriate, and end with a reason to buy now. Avoid superlatives like "best" or "revolutionary" unless you can back them up with data I provide.

What it does: Creates benefit-driven product copy that converts by connecting features to customer needs.

Why it works with Claude: The instruction to avoid unbacked superlatives aligns with Claude’s natural tendency toward honesty and precision. Claude won’t oversell — it will find compelling language that’s still accurate, which builds reader trust.

Customization tip: Paste customer reviews or testimonials into the prompt. Claude will extract the language your actual customers use and mirror it in the description.

5. Resume Bullet Points

Rewrite the following job responsibilities into strong resume bullet points using the XYZ formula (Accomplished [X] as measured by [Y] by doing [Z]). Current responsibilities: [PASTE YOUR CURRENT JOB DUTIES]. For each bullet: start with a strong action verb, include a quantifiable result (estimate reasonable metrics if I haven't provided them — flag these as estimates), and keep each bullet to one line (under 120 characters). Generate 6-8 bullets and rank them by impact.

What it does: Transforms bland job descriptions into quantified, achievement-focused resume bullets that pass ATS screening.

Why it works with Claude: Claude’s transparency shines here — it will clearly flag estimated metrics rather than fabricating numbers. The ranking instruction leverages Claude’s analytical ability to assess which achievements would most impress a hiring manager.

Customization tip: Add “Target role: [JOB TITLE]” to tailor the language to your desired position’s keywords.

6. Cover Letter

Write a cover letter for the following job posting: [PASTE FULL JOB DESCRIPTION]. My background: [PASTE RESUME OR KEY QUALIFICATIONS]. The letter should be 300-400 words, open with a specific reason I'm excited about THIS company (not generic enthusiasm), connect exactly 3 of my qualifications to their stated requirements, include one brief story or specific example that demonstrates a relevant skill, and close with confidence without being arrogant. Do not start with "I am writing to apply for."

What it does: Creates a personalized, compelling cover letter that directly maps your qualifications to the job requirements.

Why it works with Claude: Pasting the full job description uses Claude’s large context window. The negative constraint (“Do not start with…”) prevents the most common AI cover letter opener, and Claude follows these exclusions reliably.

Customization tip: Add the company’s “About Us” page text so Claude can reference specific company values and initiatives.

7. Meeting Summary

Summarize the following meeting transcript into a structured meeting summary. Transcript: [PASTE TRANSCRIPT]. Format the summary as: 1) Meeting title, date, and attendees, 2) Three-sentence executive summary, 3) Key decisions made (bulleted list), 4) Action items with owner and deadline for each, 5) Open questions or unresolved issues, 6) Next meeting date/agenda items if mentioned. If any action items are missing a clear owner or deadline, flag them explicitly.

What it does: Converts raw meeting transcripts into structured, actionable summaries with tracked action items.

Why it works with Claude: Claude’s 200K context window handles full meeting transcripts (a 1-hour meeting produces roughly 10,000-15,000 words). The instruction to flag missing owners and deadlines uses Claude’s analytical capability to catch gaps in accountability.

Customization tip: Use this with Zoom, Otter.ai, or Fireflies transcripts pasted directly. Add “Also note any off-topic tangents that consumed more than 2 minutes” to improve future meeting efficiency.

8. Press Release

Write a press release for the following announcement: [DESCRIBE THE NEWS]. Company details: [NAME, LOCATION, WHAT YOU DO]. Follow AP style and standard press release format: headline (under 80 characters), dateline, lead paragraph answering who/what/when/where/why, 2-3 body paragraphs with supporting details and context, one direct quote from [SPOKESPERSON NAME AND TITLE], boilerplate "About [COMPANY]" paragraph, and media contact information placeholder. Keep the total length under 500 words.

What it does: Produces a properly formatted, AP-style press release ready for distribution.

Why it works with Claude: Claude follows the specific formatting conventions of press releases (inverted pyramid, dateline format, boilerplate positioning) with high accuracy. Its instruction-following ensures every required element appears in the correct order.

Customization tip: Add “Include a statistic or data point in the lead paragraph” to make the release more newsworthy and quotable.

9. Newsletter

Write a weekly newsletter edition about [TOPIC/THEME]. Newsletter name: [NAME]. Audience: [DESCRIBE SUBSCRIBERS]. Structure: 1) Subject line (under 50 characters, curiosity-driven), 2) Opening hook — a surprising stat, question, or observation (2-3 sentences), 3) Main story (300 words) with a clear insight or lesson, 4) 3 curated links with one-sentence descriptions of why each matters, 5) One actionable tip readers can implement today, 6) Sign-off with a question to encourage replies. Total length: under 600 words. Tone: [SPECIFY].

What it does: Creates a complete newsletter edition with engagement-optimized structure and curated content.

Why it works with Claude: The multi-section structure tests Claude’s ability to shift between different writing modes (hook writing, analysis, curation, conversational) within one output. Claude handles these transitions smoothly without losing coherence.

Customization tip: Paste your last 2-3 newsletter editions into the prompt so Claude can match your established voice and format.

10. Speech Outline

Create a detailed outline for a [LENGTH]-minute speech about [TOPIC]. Audience: [DESCRIBE]. The outline should include: an opening hook (story, question, or startling fact — give me 2 options), a clear thesis statement, 3 main points with 2 supporting sub-points each, transition sentences between main points, one audience interaction moment (question, show of hands, or brief exercise), a memorable closing that calls back to the opening hook, and estimated time allocation per section. Mark places where I should pause, use visuals, or shift tone.

What it does: Generates a detailed, time-allocated speech outline with delivery cues and audience engagement moments.

Why it works with Claude: Claude excels at this level of structural detail. The request for time allocation per section and delivery cues asks Claude to think from the speaker’s perspective, not just the content perspective — and Claude’s nuanced reasoning handles this dual-awareness well.

Customization tip: Specify whether the speech is for a conference, team meeting, classroom, or toast — Claude will adjust formality and interaction style accordingly.

---

Business & Strategy Prompts (11–20)

Claude’s ability to maintain logical consistency across long, structured outputs makes it particularly strong for strategic analysis. A 2025 Stanford HAI report found that AI-assisted strategic planning reduced analysis time by 40% while improving the identification of competitive blind spots. These prompts help you tap into that capability.

11. SWOT Analysis

Conduct a SWOT analysis for [COMPANY/PRODUCT/IDEA]. Context: [PROVIDE RELEVANT DETAILS — industry, size, stage, competitors]. For each quadrant (Strengths, Weaknesses, Opportunities, Threats), provide exactly 5 items. Each item should be: one specific sentence (not vague), grounded in the context I provided, and actionable — meaning I can either leverage it (S/O) or mitigate it (W/T). After the four quadrants, add a "Strategic Implications" section with 3 prioritized recommendations that connect specific strengths to specific opportunities.

What it does: Produces a structured SWOT analysis with actionable strategic recommendations grounded in your specific context.

Why it works with Claude: The “exactly 5 items” constraint prevents Claude from padding weak quadrants with generic filler. The cross-referencing requirement in the Strategic Implications section leverages Claude’s ability to reason across sections of its own output.

Customization tip: Paste your company’s latest annual report or investor deck for deeper, more data-grounded analysis.

12. Business Plan Section

Write the [SPECIFIC SECTION — e.g., "Market Analysis" or "Financial Projections"] section of a business plan for [BUSINESS DESCRIPTION]. Industry: [INDUSTRY]. Target market: [DESCRIBE]. Stage: [idea/startup/growth/mature]. The section should be 800-1,000 words, include at least 3 cited data points (use recent industry reports — cite the source), present information in a format an investor or loan officer would expect, and flag any assumptions I should validate with my own research. If I haven't provided enough information for a credible section, tell me exactly what data points you need.

What it does: Creates an investor-grade business plan section with real data and transparent assumptions.

Why it works with Claude: Claude’s willingness to say “I need more information” and flag assumptions is a product of its Constitutional AI training, which prioritizes honesty over helpfulness-at-all-costs. This produces more credible business documents.

Customization tip: Run this prompt for each section sequentially, pasting the previous sections into each new prompt so Claude maintains consistency across the full plan.

13. Competitive Analysis

Create a competitive analysis comparing [YOUR PRODUCT/SERVICE] against these competitors: [LIST 3-5 COMPETITORS]. Analyze each competitor across these dimensions: 1) Core value proposition, 2) Pricing model and price points, 3) Target customer segment, 4) Key strengths, 5) Key weaknesses, 6) Market positioning. Present this as a comparison table, then write a 200-word "Competitive Advantage Summary" identifying the gaps in the market that [YOUR PRODUCT] can exploit. Be brutally honest about where competitors are stronger — I need real intelligence, not validation.

What it does: Delivers an honest competitive landscape analysis with a table format for quick scanning and strategic gaps identification.

Why it works with Claude: The “be brutally honest” instruction works because Claude’s training doesn’t have a strong sycophancy bias. It will give you uncomfortable truths about competitive weaknesses when explicitly asked.

Customization tip: Add competitor website URLs and product pages for Claude to analyze directly within its context window.

14. Customer Persona

Create a detailed customer persona for [PRODUCT/SERVICE]. What I know about my customers: [SHARE ANY DATA — demographics, behaviors, feedback, purchase patterns]. Build the persona with: Name and demographic snapshot, Job title and daily responsibilities, 3 primary goals related to my product category, 3 frustrations/pain points with current solutions, Buying behavior (how they research, who influences them, what triggers a purchase), Preferred communication channels, A direct quote that captures their mindset (written as if from a real interview), and One "surprising insight" — something about this persona that's counterintuitive but strategically valuable.

What it does: Builds a rich, actionable customer persona that goes beyond demographics into motivations and decision-making patterns.

Why it works with Claude: The “surprising insight” element leverages Claude’s reasoning ability to synthesize non-obvious patterns from the data you provide. Claude consistently produces counterintuitive observations that challenge assumptions.

Customization tip: Paste actual customer support tickets, survey responses, or reviews to give Claude real voice-of-customer data to work from.

15. Marketing Strategy

Develop a 90-day marketing strategy for [PRODUCT/SERVICE]. Budget: [AMOUNT or "limited/moderate/substantial"]. Current situation: [DESCRIBE — new launch, growth stage, rebrand, etc.]. Target audience: [DESCRIBE]. The strategy should include: 3 specific, measurable goals with KPIs, channel recommendations (ranked by expected ROI for my budget level), a week-by-week timeline for the first 30 days with specific deliverables, content themes for each month, 2 quick wins I can execute this week, and 1 experimental tactic to test. For each channel, estimate the time commitment required so I can assess feasibility.

What it does: Creates a time-bound, budget-aware marketing plan with measurable goals and a concrete execution timeline.

Why it works with Claude: The week-by-week breakdown tests Claude’s ability to think sequentially and realistically about resource allocation. Claude handles time-based planning well because it can reason about dependencies between tasks.

Customization tip: Add “My team consists of [X people] with these skills: [LIST]” to get feasibility-adjusted recommendations.

16. Pricing Strategy

Help me develop a pricing strategy for [PRODUCT/SERVICE]. Current pricing: [IF ANY]. Costs: [YOUR COSTS/MARGINS]. Competitor pricing: [LIST WHAT YOU KNOW]. Target market: [DESCRIBE]. Analyze 3 viable pricing models for my situation (e.g., freemium, tiered, usage-based, flat-rate, per-seat). For each model: explain the pros and cons for MY specific situation, recommend specific price points with reasoning, project the impact on customer acquisition vs. revenue per customer, and identify the key metric I should track to know if it's working. End with your top recommendation and why.

What it does: Evaluates multiple pricing models against your specific business context and recommends the optimal approach with data-driven reasoning.

Why it works with Claude: Claude handles multi-option analysis without losing track of the evaluation criteria across all options. Its structured reasoning ensures each pricing model gets equally rigorous treatment.

Customization tip: Include your current conversion rate and customer lifetime value if known — Claude will factor these into its ROI projections.

17. Job Description

Write a job description for a [JOB TITLE] role at [COMPANY TYPE/SIZE]. This role is responsible for: [KEY RESPONSIBILITIES]. Reports to: [ROLE]. Team size: [X]. Structure the description with: a 2-3 sentence role summary that sells the opportunity (not just lists duties), "What You'll Do" section with 6-8 specific responsibilities (action-verb format), "What You'll Bring" section split into "Required" (5-6 real requirements) and "Nice to Have" (3-4 genuine preferences), salary range: [RANGE or "competitive"], and one section on culture/benefits that feels authentic, not corporate-generic. Avoid exclusionary language and unnecessary requirements that could limit the candidate pool.

What it does: Produces an inclusive, compelling job description that attracts qualified candidates by balancing requirements with appeal.

Why it works with Claude: Claude’s CAI training makes it naturally attentive to exclusionary language (unnecessary degree requirements, gendered phrasing, inflated experience demands). The instruction to avoid these practices aligns with Claude’s built-in principles.

Customization tip: Paste a top-performer’s actual work output or achievements, and Claude will reverse-engineer the skills and traits that matter most for the role.

18. Performance Review

Help me write a performance review for [EMPLOYEE NAME/ROLE]. Review period: [TIMEFRAME]. Key accomplishments: [LIST]. Areas for development: [LIST]. Overall rating: [EXCEEDS/MEETS/BELOW EXPECTATIONS]. Write the review with: an opening summary paragraph that sets a constructive tone, 3-4 specific achievement paragraphs (each tied to a business outcome, not just activity), 2 development areas framed as growth opportunities with specific, actionable suggestions, 2-3 goals for the next review period that are SMART (Specific, Measurable, Achievable, Relevant, Time-bound), and a closing paragraph. The tone should be honest, specific, and growth-oriented — not corporate fluff. Every piece of feedback should include a specific example.

What it does: Creates a thorough, balanced performance review with specific examples and actionable development goals.

Why it works with Claude: Claude handles the nuance of development feedback well — framing areas for improvement constructively without sugarcoating. Its training specifically optimizes for being helpful and honest simultaneously.

Customization tip: Run a second prompt: “Now write a self-review from the employee’s perspective based on the same facts” to prepare for a two-way conversation.

19. Project Brief

Create a project brief for [PROJECT NAME]. Objective: [WHAT WE'RE TRYING TO ACHIEVE]. Stakeholders: [LIST]. Timeline: [DURATION]. Budget: [AMOUNT/RANGE]. The brief should include: Problem statement (what's broken or missing today — 2-3 sentences), Project objectives (3-5 SMART goals), Scope (what's included AND what's explicitly excluded), Key deliverables with target dates, Success metrics (how we'll know this worked), Risks (top 3) with mitigation strategies for each, Resource requirements, and Decision rights (who approves what). Keep it under 2 pages. Use a professional but readable tone — this will be shared with both executives and the implementation team.

What it does: Produces a complete project brief that aligns stakeholders on scope, success criteria, and decision-making before work begins.

Why it works with Claude: The “explicitly excluded” scope requirement leverages Claude’s ability to think about what’s NOT included — a critical skill for preventing scope creep that Claude handles better than models that default to being maximally helpful.

Customization tip: Add “The main source of disagreement among stakeholders is [X]” and Claude will address that tension directly in the brief.

20. Sales Email Sequence

Write a 5-email sales sequence for [PRODUCT/SERVICE]. Target buyer: [DESCRIBE — role, company size, pain points]. Sequence structure: Email 1 (Day 1): Cold intro — lead with a pain point, not your product. Under 100 words. Email 2 (Day 3): Value-add — share a useful insight or resource. No hard sell. Email 3 (Day 7): Social proof — case study or specific result. Include one number. Email 4 (Day 10): Direct offer with clear CTA and deadline/urgency element. Email 5 (Day 14): Breakup email — final touch, respectful close, leave the door open. For each email: write the subject line, body, and CTA. Subject lines should be under 40 characters and avoid spam trigger words. All emails should feel like they come from a real person, not a marketing system.

What it does: Creates a complete 5-email outbound sales sequence with strategic escalation from value-first to direct offer.

Why it works with Claude: Claude maintains consistent voice and strategic progression across all 5 emails. Each email builds on the previous one logically. The “feel like a real person” constraint produces natural language rather than marketing-speak.

Customization tip: After generating, ask: “Now write 3 alternative subject lines for each email, optimized for open rate” to A/B test.

---

Research & Analysis Prompts (21–30)

Claude’s reasoning capabilities and large context window make it powerful for research tasks. According to Anthropic’s technical documentation, Claude 3.5 Sonnet achieves state-of-the-art performance on reasoning benchmarks including GPQA (graduate-level science questions), demonstrating genuine analytical capability beyond pattern matching.

21. Literature Review

I'm researching [TOPIC] for a [paper/report/presentation]. Here are [X] sources I've gathered: [PASTE ABSTRACTS, EXCERPTS, OR FULL TEXTS]. Create a literature review that: groups the sources thematically (not chronologically), identifies 3-4 key themes or debates in the literature, notes where sources agree and where they contradict each other, identifies gaps in the current research that future work could address, and maintains an objective academic tone. Use in-text citations in [APA/MLA/Chicago] format. End with a synthesis paragraph that connects the themes to my research question: [STATE YOUR QUESTION].

What it does: Synthesizes multiple sources into a cohesive, thematically organized literature review with proper academic formatting.

Why it works with Claude: Claude’s 200K context window means you can paste entire papers, not just abstracts. Claude excels at identifying contradictions between sources — a task that requires holding multiple viewpoints simultaneously, which its multi-step reasoning handles well.

Customization tip: For systematic reviews, add: “Also create a table summarizing each source’s methodology, sample size, key finding, and limitation.”

22. Data Analysis Prompt

Analyze the following dataset: [PASTE DATA OR DESCRIBE THE DATASET]. I want to understand: [STATE YOUR SPECIFIC QUESTIONS — e.g., "What factors most strongly predict customer churn?"]. Please: 1) Describe the dataset (size, variables, data types, any obvious quality issues), 2) Identify the 3 most important patterns or trends, 3) Note any outliers or anomalies that warrant investigation, 4) Suggest 2-3 statistical tests or analyses that would answer my questions (explain why each is appropriate), 5) Summarize findings in plain English — as if presenting to a non-technical stakeholder, 6) List 3 follow-up questions the data raises. If the data is insufficient to draw reliable conclusions, say so explicitly.

What it does: Provides a structured data analysis with statistical recommendations, plain-language findings, and honest assessment of data limitations.

Why it works with Claude: Claude’s willingness to say “this data is insufficient” prevents the common AI trap of finding patterns in noise. The dual-audience instruction (technical analysis + plain English summary) leverages Claude’s ability to code-switch between registers.

Customization tip: Paste CSV data directly into the prompt. Claude can parse tabular data and analyze it without external tools for datasets under roughly 50K tokens.

23. Market Research

Conduct market research analysis for [PRODUCT/SERVICE/INDUSTRY]. What I need to understand: market size and growth trajectory, key customer segments and their needs, competitive landscape (major players and market share distribution), barriers to entry, current trends shaping the market, and regulatory factors if applicable. Structure the output as a market research brief that I could present to investors or leadership. Include specific numbers and cite sources where possible. Where you're uncertain about a data point, give your best estimate and mark it clearly as "[ESTIMATE]" so I know what to verify independently. End with 3 strategic implications for my business.

What it does: Creates an investor-ready market research brief with data, competitive analysis, and strategic implications.

Why it works with Claude: The [ESTIMATE] flagging instruction works exceptionally well with Claude because of its commitment to transparency. Claude will genuinely distinguish between data it’s confident about and figures it’s approximating.

Customization tip: Provide your target geography (“US only” or “Global excluding China”) for more focused market sizing.

24. Pros/Cons Analysis

I need to decide between these options: [LIST 2-4 OPTIONS WITH BRIEF DESCRIPTIONS]. My priorities are (ranked): [LIST WHAT MATTERS MOST TO YOU]. My constraints are: [BUDGET, TIMELINE, TEAM SIZE, etc.]. For each option, provide: 5 specific pros (tied to my stated priorities), 5 specific cons (including hidden costs or risks I might not have considered), a suitability score from 1-10 for my specific situation, and one "deal-breaker" scenario — a situation where this option would clearly be the wrong choice. Then provide your recommendation with reasoning. Be direct — I need a clear recommendation, not "it depends."

What it does: Delivers a structured decision-analysis framework with weighted evaluation against your specific priorities and constraints.

Why it works with Claude: The “be direct — I need a clear recommendation” instruction overcomes the common AI tendency to hedge. Claude will commit to a recommendation while showing its reasoning, which is more useful than a non-answer.

Customization tip: Add “My risk tolerance is [low/medium/high]” to shift the analysis toward safer or more aggressive options accordingly.

25. Executive Summary

Write an executive summary of the following document: [PASTE FULL DOCUMENT]. The summary should be [250/500/750] words and structured for [AUDIENCE — e.g., C-suite, board of directors, team leads]. Include: the core problem or opportunity (1-2 sentences), key findings or recommendations (bulleted — no more than 5), financial impact or resource implications if applicable, recommended next steps with timeline, and the single most important thing the reader should remember. Write it so that someone who reads ONLY this summary can make an informed decision without reading the full document.

What it does: Condenses lengthy documents into decision-ready executive summaries tailored to specific audience levels.

Why it works with Claude: Pasting the full document (not a summary of a summary) into Claude’s large context window produces dramatically better executive summaries. Claude can identify what matters most rather than just summarizing sequentially.

Customization tip: For board decks, add: “Include a risk section and frame recommendations as decision points with clear options.”

26. Fact-Checking

Fact-check the following text: [PASTE CONTENT TO CHECK]. For each factual claim in the text: 1) Identify the claim, 2) Rate your confidence in its accuracy: CONFIRMED (you're very confident it's correct), LIKELY (probably correct but you have some uncertainty), UNVERIFIABLE (you can't confirm or deny), LIKELY INCORRECT (probably wrong), or INCORRECT (you're confident it's wrong), 3) Provide the correct information if the claim is wrong, 4) Note the basis for your assessment. Also flag any misleading framing — claims that are technically true but presented in a way that could lead to wrong conclusions. Summarize: how many claims total, how many confirmed, how many flagged.

What it does: Systematically evaluates factual claims with a confidence-rated framework that distinguishes between certainty levels.

Why it works with Claude: The graduated confidence scale (instead of binary true/false) plays to Claude’s calibration — Anthropic’s research shows Claude’s confidence levels correlate well with actual accuracy. The “misleading framing” check uses Claude’s nuanced reasoning to catch deception-by-omission.

Customization tip: For scientific or medical claims, add: “Apply a higher evidence standard — require peer-reviewed sources for CONFIRMED rating.”

27. Trend Analysis

Analyze the current trends in [INDUSTRY/TOPIC] as of early 2026. Structure your analysis as: 1) Top 5 trends, ranked by likely impact over the next 12 months — for each trend, provide evidence it's real (not just hype), who's affected most, and what it means practically for [MY ROLE/BUSINESS], 2) One "contrarian take" — a widely discussed trend that you think is overblown, with reasoning, 3) One "under-the-radar" trend that isn't getting enough attention, 4) A 3-sentence "so what" summary I could share with my team. Base this on publicly available information up to your training cutoff. Be specific — I want named companies, real products, and actual developments, not generic trend labels.

What it does: Delivers a ranked, evidence-based trend analysis with both consensus and contrarian perspectives.

Why it works with Claude: The “contrarian take” instruction is where Claude shines — it’s willing to push back against conventional wisdom when the evidence supports it, which is rare for AI models that tend toward consensus validation.

Customization tip: Add “My company is in [SECTOR] with [SIZE] employees” so Claude can filter trends through your specific relevance lens.

28. Survey Question Generator

Generate a survey to measure [WHAT YOU WANT TO LEARN] from [TARGET RESPONDENTS]. Create 15 questions following these rules: Mix of question types (Likert scale, multiple choice, open-ended, ranking), Start with easy/non-threatening questions and progress to more sensitive ones, Include 1 attention-check question to filter low-quality responses, No double-barreled questions (one concept per question), No leading questions (neutral wording), Include skip logic recommendations (e.g., "If Q3 = No, skip to Q7"). For each question, note what insight it provides and how I'd use the response data. Estimate completion time. End with 3 questions I should NOT ask (and why) — common mistakes for surveys on this topic.

What it does: Creates a methodologically sound survey with proper question design, flow logic, and quality controls.

Why it works with Claude: The instruction to note what each question measures forces Claude to think instrumentally about survey design rather than just generating questions. The “questions NOT to ask” section leverages Claude’s ability to reason about what makes a bad question.

Customization tip: Specify your survey tool (Google Forms, Typeform, SurveyMonkey) and Claude will note which question types are supported on that platform.

29. Report Outline

Create a detailed outline for a report on [TOPIC]. Report type: [annual report/research report/audit report/white paper/internal memo]. Target audience: [DESCRIBE]. Expected length: [PAGES/WORDS]. The outline should include: section titles and sub-sections (3 levels deep), a 1-2 sentence description of what each section should contain, suggested data visualizations or tables for each section (describe what the chart would show), the key question each section should answer, and recommended sources or data I'll need to gather. Organize the sections in a logical flow that builds an argument toward [THE MAIN CONCLUSION OR RECOMMENDATION].

What it does: Produces a comprehensive report blueprint with section descriptions, data visualization suggestions, and research requirements.

Why it works with Claude: Claude’s ability to think three levels deep in hierarchical structure while maintaining logical flow between sections is a direct benefit of its training on complex, long-form content. The argument-building instruction ensures coherence.

Customization tip: Add “The report needs to convince [SKEPTICAL AUDIENCE] of [CONCLUSION]” to sharpen the argumentative structure.

30. Source Evaluation

Evaluate the following source for credibility and usefulness for my research on [TOPIC]: [PASTE THE SOURCE TEXT OR DESCRIBE IT WITH PUBLICATION, AUTHOR, DATE]. Assess the source on: 1) Authority — author credentials and publication reputation, 2) Accuracy — are claims verifiable? Any obvious errors?, 3) Objectivity — signs of bias, funding sources, or conflicts of interest, 4) Currency — is the information current enough for my purpose?, 5) Coverage — how thoroughly does it address the topic?, 6) Methodology — if it's research, is the methodology sound? Give an overall rating: PRIMARY SOURCE (build arguments on this), SUPPORTING SOURCE (use to supplement), BACKGROUND ONLY (provides context but don't cite directly), or UNRELIABLE (don't use). Explain your rating in 2-3 sentences.

What it does: Provides a structured credibility assessment of research sources using established evaluation criteria.

Why it works with Claude: Claude’s training on Constitutional AI principles makes it particularly good at detecting bias and evaluating objectivity. It will flag subtle issues like cherry-picked statistics or conflicts of interest that might not be immediately obvious.

Customization tip: Batch multiple sources in one prompt: “Evaluate these 5 sources and rank them by reliability for my specific research question.”

---

Coding & Technical Prompts (31–40)

Claude consistently ranks among the top models for coding tasks. On the SWE-bench benchmark (which tests real-world software engineering tasks from GitHub issues), Claude 3.5 Sonnet achieved a 49% resolution rate as of its October 2024 release — the highest at the time of testing. For a deeper look at Claude’s coding capabilities, see our Claude Code beginner’s guide.

31. Code Review

Review the following code for quality, bugs, and improvements: [PASTE CODE]. Language: [LANGUAGE]. Context: [WHAT THE CODE DOES]. Review for: 1) Bugs or logical errors (critical — flag these first), 2) Security vulnerabilities (SQL injection, XSS, etc.), 3) Performance issues (O(n²) loops, unnecessary database calls, memory leaks), 4) Code style and readability (naming, structure, comments), 5) Edge cases not handled, 6) Suggestions for improvement (explain why, not just what). Rate each finding as CRITICAL, IMPORTANT, or NICE-TO-HAVE. Show corrected code for any CRITICAL findings. Don't just say "looks good" — find at least 3 concrete suggestions even if the code is solid.

What it does: Performs a comprehensive code review with severity-rated findings, security checks, and corrected code for critical issues.

Why it works with Claude: Claude’s ability to hold entire codebases in context means it can spot cross-function issues that file-by-file review would miss. The “don’t just say looks good” instruction prevents sycophantic code review — a common issue with AI assistants.

Customization tip: Add your team’s style guide or linting rules, and Claude will review against your specific standards, not generic best practices.

32. Bug Fix

I have a bug in my code. Here's the code: [PASTE CODE]. Expected behavior: [WHAT SHOULD HAPPEN]. Actual behavior: [WHAT'S HAPPENING INSTEAD]. Error message (if any): [PASTE ERROR]. What I've already tried: [LIST DEBUGGING STEPS]. Please: 1) Identify the root cause (not just the symptom), 2) Explain WHY the bug occurs (help me understand, not just fix), 3) Provide the corrected code, 4) Show me how to verify the fix works, 5) Suggest a test case that would have caught this bug, 6) Note if this bug pattern could exist elsewhere in similar code. If you need more context (e.g., other files, configuration, dependencies), tell me exactly what you need.

What it does: Diagnoses and fixes bugs with root-cause analysis, educational explanation, and preventive testing recommendations.

Why it works with Claude: The “explain WHY” instruction leverages Claude’s teaching capability. The “what I’ve already tried” section prevents Claude from suggesting solutions you’ve already ruled out — a frustrating loop with less context-aware models.

Customization tip: Include your stack details (framework versions, OS, runtime) — version-specific bugs are extremely common and context makes diagnosis much more accurate.

33. Refactor

Refactor the following code to improve [readability/performance/maintainability/testability — pick one primary goal]: [PASTE CODE]. Language: [LANGUAGE]. Context: [WHAT IT DOES AND WHERE IT FITS IN THE CODEBASE]. Constraints: [ANY REQUIREMENTS — e.g., "must maintain backward compatibility," "can't add new dependencies," "must keep the public API unchanged"]. Please: 1) Show the refactored code, 2) Explain each change and why it improves the code, 3) List any trade-offs introduced by the refactoring, 4) Confirm the refactored version produces identical outputs for all inputs, 5) If the refactoring is large, show it in stages so I can apply incrementally. Do NOT refactor just to make the code "look different." Every change should have a clear benefit.

What it does: Produces a principled code refactoring with clear rationale, trade-off analysis, and incremental implementation guidance.

Why it works with Claude: The “do NOT refactor just to look different” instruction prevents the over-refactoring problem common with AI coding assistants. Claude respects constraints like backward compatibility because its instruction-following is precise.

Customization tip: Paste your team’s coding standards document and ask Claude to refactor toward those specific standards.

34. Write Tests

Write comprehensive tests for the following code: [PASTE CODE]. Language: [LANGUAGE]. Testing framework: [Jest/pytest/JUnit/etc.]. The test suite should include: happy path tests (expected inputs → expected outputs), edge case tests (empty inputs, boundary values, maximum sizes), error case tests (invalid inputs, network failures, permission errors), at least one integration test if the code interacts with external services, and clear test names that describe the expected behavior (not "test1", "test2"). Aim for [X]% code coverage. For each test, add a one-line comment explaining what scenario it covers. If the code is hard to test, suggest refactoring changes that would improve testability.

What it does: Generates a thorough test suite with edge cases, error handling, and clear documentation of test purpose.

Why it works with Claude: Claude’s systematic thinking produces edge cases that developers often miss. The instruction to suggest testability refactoring leverages Claude’s ability to reason about code structure holistically.

Customization tip: Add “My CI pipeline runs [TOOL] — ensure test output format is compatible” for immediate integration.

35. Explain Code

Explain the following code to me as if I'm a [beginner/intermediate/senior] developer: [PASTE CODE]. I want to understand: 1) What does this code do overall? (2-3 sentence summary), 2) Walk through the code line-by-line (or section-by-section for longer code), explaining the purpose and logic, 3) What design patterns or techniques does it use? (name them), 4) What are the inputs and outputs?, 5) Where could this code fail, and how does it handle (or not handle) failures?, 6) If you were explaining this to my team, what's the one thing you'd emphasize? Use analogies where they help, but don't oversimplify technical concepts — I want to actually learn.

What it does: Provides a multi-level code explanation from high-level summary through line-by-line walkthrough to design pattern identification.

Why it works with Claude: The skill-level adjustment (“as if I’m a [X] developer”) works because Claude genuinely shifts its explanation depth and vocabulary. It won’t condescend to seniors or overwhelm beginners. The analogies instruction produces surprisingly apt comparisons.

Customization tip: Specify your background: “I know Python well but I’m new to Go” — Claude will bridge from familiar concepts to unfamiliar ones.

36. API Integration

Write code to integrate with [API NAME — e.g., Stripe, Twilio, SendGrid]. Language: [LANGUAGE]. Framework: [IF APPLICABLE]. I need to: [DESCRIBE WHAT YOU WANT TO DO — e.g., "process payments," "send SMS notifications," "manage email lists"]. Include: 1) Environment setup (dependencies, API key configuration — never hardcode secrets), 2) Authentication setup, 3) The main integration code with error handling, 4) Retry logic for transient failures (with exponential backoff), 5) Logging for debugging, 6) A simple test to verify the integration works (using the API's test/sandbox mode if available), 7) Rate limiting considerations. Add comments explaining any non-obvious API quirks or gotchas you know about.

What it does: Produces production-ready API integration code with security best practices, error handling, and operational considerations.

Why it works with Claude: Claude’s training data includes extensive API documentation, so it knows common gotchas (Stripe’s idempotency keys, Twilio’s message segment pricing, etc.). The explicit “never hardcode secrets” instruction aligns with Claude’s security-conscious defaults.

Customization tip: Paste the API’s rate limit documentation and ask Claude to build the rate limiter to match exactly.

37. Database Query

Write a [SQL/MongoDB/etc.] query for the following: [DESCRIBE WHAT DATA YOU NEED]. Database structure: [DESCRIBE YOUR TABLES/COLLECTIONS — or paste your schema]. Requirements: The query should be optimized for performance on a table with [APPROXIMATE ROW COUNT] rows. Include appropriate indexes if needed (suggest CREATE INDEX statements). Add comments explaining the logic of any complex joins, subqueries, or window functions. If there are multiple ways to write this query, show the most performant option and briefly explain why it's faster. Also show the execution plan analysis approach (EXPLAIN/EXPLAIN ANALYZE) so I can verify performance.

What it does: Creates performance-optimized database queries with indexing recommendations, inline documentation, and performance verification guidance.

Why it works with Claude: Providing your actual schema (not a description of it) lets Claude write queries against real column names and data types. Claude’s SQL optimization suggestions are grounded in actual query planner behavior, not just theoretical best practices.

Customization tip: Add “Database: [PostgreSQL 16/MySQL 8/etc.]” — SQL dialects differ significantly and Claude will use version-specific features and syntax.

38. Documentation

Write documentation for the following code: [PASTE CODE]. Documentation type: [API docs/README/inline comments/user guide — pick one]. Audience: [developers on my team/open source contributors/non-technical users]. Include: 1) Overview — what this code does and why it exists (not just what, but why), 2) Setup/installation instructions (prerequisites, dependencies, environment variables), 3) Usage examples — at least 3, progressing from simple to advanced, 4) API reference — every public function/method with parameters, return values, and exceptions, 5) Common pitfalls or FAQ (at least 3 items), 6) Troubleshooting section for the most likely errors. Write for someone who's encountering this code for the first time. If the code has poor naming or structure that makes documentation difficult, note that as a suggestion.

What it does: Generates comprehensive, audience-appropriate documentation with practical examples and troubleshooting guidance.

Why it works with Claude: The “why it exists, not just what” instruction produces documentation that helps developers understand intent, which is the most common gap in code documentation. Claude’s progressive examples (simple to advanced) follow good pedagogical structure.

Customization tip: Specify your documentation tool (JSDoc, Sphinx, Swagger) and Claude will format output in that tool’s expected syntax.

39. Regex Generator

Create a regular expression that matches [DESCRIBE THE PATTERN YOU NEED TO MATCH]. Here are example strings that SHOULD match: [LIST 3-5 EXAMPLES]. Here are example strings that should NOT match: [LIST 3-5 EXAMPLES]. Language/engine: [JavaScript/Python/PCRE/etc.]. Please: 1) Provide the regex pattern, 2) Explain each part of the pattern in plain English, 3) Show test cases proving it works (matches what it should, rejects what it shouldn't), 4) Note any edge cases that might cause unexpected matches or misses, 5) If the regex is complex, provide a simpler alternative that covers 90% of cases (with a note about what the 10% misses are). Performance note: will this pattern cause catastrophic backtracking on long strings?

What it does: Builds tested, documented regular expressions with plain-English explanations and edge case analysis.

Why it works with Claude: Providing both positive and negative examples gives Claude clear boundaries. The catastrophic backtracking check is crucial — Claude will identify ReDoS-vulnerable patterns, which is a security concern many developers miss.

Customization tip: Always include the target engine — JavaScript and Python regex engines handle lookaheads and Unicode differently.

40. Architecture Review

Review the following system architecture for [PROJECT/APPLICATION]: [DESCRIBE OR PASTE DIAGRAM/DESCRIPTION — include components, data flows, technology choices, and scale requirements]. Evaluate: 1) Does the architecture meet the stated requirements? (identify any gaps), 2) Scalability — where are the bottlenecks as traffic grows 10x and 100x?, 3) Reliability — single points of failure?, 4) Security — data flow security, authentication/authorization design, 5) Cost efficiency — are there overengineered components for the current scale?, 6) Operational complexity — is this reasonable for a team of [SIZE]? Rate each area GREEN (solid) / YELLOW (workable but has risks) / RED (needs attention before production). Provide 3 prioritized recommendations.

What it does: Delivers a comprehensive architecture review with traffic-light ratings and prioritized recommendations tailored to your team’s capacity.

Why it works with Claude: The scaling question (10x, 100x) forces Claude to think about architecture in terms of concrete growth scenarios, not abstract principles. Claude’s multi-dimensional analysis (scale, reliability, security, cost, operations) mirrors how senior architects actually evaluate systems.

Customization tip: Include your current monthly infrastructure cost. Claude can identify savings opportunities by comparing your architecture to more cost-efficient alternatives.

---

Education & Learning Prompts (41–45)

Claude’s ability to adjust explanation depth and use analogies makes it an effective learning companion. These five prompts cover the most common educational use cases, from concept explanation to full lesson planning.

41. Explain Concept

Explain [CONCEPT] to me. My current understanding: [DESCRIBE WHAT YOU ALREADY KNOW, EVEN IF IT'S "NOTHING"]. My background: [RELEVANT CONTEXT — e.g., "I'm a marketing manager," "I have a biology degree," "I'm 16"]. Explain it in 3 layers: 1) The 30-second version — one paragraph I could repeat to someone at dinner, 2) The 5-minute version — enough to understand how it works and why it matters, with an analogy from everyday life, 3) The deep dive — technical details, nuances, common misconceptions, and how experts actually think about this concept. After all three layers, give me 3 questions I should be able to answer if I truly understand this concept. If my stated current understanding contains any misconceptions, correct them gently but directly.

What it does: Provides a layered explanation at three depth levels, corrects existing misconceptions, and tests understanding with verification questions.

Why it works with Claude: The three-layer approach prevents the common AI problem of defaulting to either too-simple or too-technical explanations. Claude genuinely adjusts between layers. The misconception correction leverages Claude’s honesty — it won’t politely ignore wrong understanding.

Customization tip: Add “I learn best through [visual descriptions/step-by-step processes/real-world examples/mathematical formulas]” to match your learning style.

42. Study Guide

Create a study guide for [SUBJECT/EXAM/TOPIC]. My current knowledge level: [BEGINNER/INTERMEDIATE/ADVANCED]. Time available: [HOW LONG UNTIL THE EXAM OR DEADLINE]. The study guide should include: 1) The 10 most important concepts I must understand (ranked by exam importance/real-world relevance), 2) For each concept: a 2-3 sentence summary, one memory aid (mnemonic, analogy, or visual), and one practice question, 3) A study schedule that fits my timeframe — which topics to study first and how much time per topic, 4) 5 common mistakes students make on this material, 5) 3 "If you only have 1 hour" priorities, 6) Recommended study techniques for this specific type of material (not generic study tips). If this material is cumulative (Topic B requires understanding Topic A), note the prerequisites clearly.

What it does: Builds a prioritized, time-aware study plan with memory aids, practice questions, and study technique recommendations.

Why it works with Claude: Claude’s ability to rank by importance (not just list) and create a realistic study schedule shows its capacity for educational planning. The prerequisite mapping leverages Claude’s understanding of how knowledge builds sequentially.

Customization tip: Paste your syllabus or textbook table of contents for a study guide perfectly mapped to your course structure.

43. Quiz Generator

Create a quiz on [TOPIC] to test my understanding. Difficulty: [EASY/MEDIUM/HARD/MIXED]. Generate 15 questions: 5 multiple choice (with 4 options each — make the wrong answers plausible, not obviously wrong), 5 true/false (include the reasoning for each answer), 3 short answer, 2 scenario-based application questions (present a situation and ask me to apply the concept). For each question: indicate the difficulty level and which sub-topic it tests. Provide all answers in a separate section at the end (not inline — I want to attempt the quiz first). Include a scoring guide: what score indicates mastery vs. what score means I need more study.

What it does: Creates a comprehensive quiz with multiple question types, plausible distractors, and a scoring framework for self-assessment.

Why it works with Claude: The “make wrong answers plausible” instruction is critical — many AI quiz generators create obviously wrong options. Claude’s nuanced understanding produces distractors that test real comprehension. The separated answer key respects the learning process.

Customization tip: After taking the quiz, paste your answers and ask: “Grade my quiz and explain every question I got wrong. What does my error pattern tell you about my understanding gaps?”

44. Lesson Plan

Create a lesson plan for teaching [TOPIC] to [AUDIENCE — age group and knowledge level]. Duration: [LENGTH OF LESSON]. Learning objectives: by the end, students should be able to [LIST 2-3 SPECIFIC, MEASURABLE OUTCOMES]. Structure the plan with: 1) Hook/opening activity (5 min) — something that creates curiosity, 2) Direct instruction (explain the concept — include the key points and teaching sequence), 3) Guided practice (an activity where students apply the concept with support), 4) Independent practice (an activity where students apply it alone), 5) Assessment — how to check if learning objectives were met, 6) Extension activity for fast finishers. Include specific questions to ask at each stage (not just "discuss"). Timing for each section. Note common student misconceptions about this topic and how to address them.

What it does: Produces a complete, classroom-ready lesson plan with engagement strategies, assessment tools, and misconception-aware teaching notes.

Why it works with Claude: Claude’s understanding of pedagogical sequence (I Do, We Do, You Do) produces lesson plans with proper scaffolding. The misconception awareness is particularly strong — Claude draws on its training data of educational research to predict where students typically struggle.

Customization tip: Add “My classroom has [X] students and [LIST AVAILABLE RESOURCES — technology, manipulatives, etc.]” for logistically feasible activities.

45. ELI5 Explainer

Explain [COMPLEX TOPIC] like I'm 5 years old. Rules: Use only words a 5-year-old would know (if you must use a big word, immediately define it in simple terms), use a concrete analogy from a child's world (playground, toys, food, family, animals), keep it under 200 words, make it accurate — don't sacrifice correctness for simplicity. After the ELI5 version, give me a "Now that you're 15" version (300 words, more nuance, some technical terms introduced gently). Then give me the "adult expert" version (300 words, full complexity, proper terminology). This should feel like zooming in on a map — same territory, increasing detail.

What it does: Provides three zoom levels of explanation from kindergarten-simple to expert-level, using the same core analogy thread.

Why it works with Claude: Claude’s analogies are remarkably apt at the 5-year-old level without being condescending or inaccurate. The “don’t sacrifice correctness for simplicity” instruction prevents Claude from oversimplifying to the point of being wrong — a trap many AI models fall into when asked to simplify.

Customization tip: For social sharing, the ELI5 version often makes the best content. Ask Claude to convert it into a social media post or short video script.

---

Creative & Personal Prompts (46–50)

Claude’s creativity comes from a different place than models optimized purely for engagement. Its Constitutional AI training produces creative outputs that are thoughtful and substantive rather than flashy and superficial. These five prompts tap into that for personal and creative projects.

46. Story Outline

Create a detailed story outline for a [GENRE] [short story/novel/screenplay]. Premise: [YOUR BASIC IDEA — even if it's rough]. The outline should include: 1) One-sentence logline, 2) Three-act structure with major plot points for each act, 3) Main character profile — name, defining trait, internal conflict, external goal, and a specific flaw that drives the plot, 4) 2-3 supporting characters with their relationship to the protagonist and their own motivations, 5) The central conflict and its escalation beats (at least 4 stages of increasing stakes), 6) The climax — what choice does the protagonist face?, 7) The resolution — how has the character changed?, 8) 3 potential opening scenes (give me options). Theme: what is this story really about beneath the plot? If my premise has structural weaknesses, tell me — I'd rather fix them now than in draft 3.

What it does: Generates a comprehensive story outline with character arcs, escalating conflict, and thematic depth — plus honest structural feedback.

Why it works with Claude: Claude’s willingness to critique your premise (per the last instruction) is unusual for AI models that default to enthusiastic support. Claude will identify plot holes, motivation gaps, and structural issues because its training values honesty alongside helpfulness.

Customization tip: Specify “tone of [REFERENCE WORK — e.g., ‘tone of The Martian’ or ‘atmosphere of Blade Runner’]” for targeted creative direction.

47. Recipe Modification

Modify this recipe to fit my dietary needs: [PASTE RECIPE OR DESCRIBE THE DISH]. My dietary requirements: [LIST — e.g., "gluten-free," "reduce sodium by 50%," "make it vegan," "under 500 calories per serving"]. Please: 1) List every substitution with the original ingredient → replacement, 2) Explain why each substitution works (chemistry/texture/flavor), 3) Note any technique changes needed (cooking time, temperature, method), 4) Flag if any substitution significantly changes the taste or texture — set my expectations, 5) Provide adjusted nutritional estimate per serving (calories, protein, key macros), 6) Suggest one optional addition that would elevate the modified version. If the dish fundamentally can't be made well with these restrictions (e.g., "make a soufflé without eggs"), be honest about that and suggest an alternative dish that achieves the same flavor profile.

What it does: Adapts recipes to dietary restrictions with scientific explanations for substitutions and honest expectations about the result.

Why it works with Claude: Claude’s honesty means it will tell you when a modification produces an inferior dish rather than pretending cauliflower tastes like bread. The food science explanations leverage Claude’s knowledge base to make you a better cook, not just follow instructions.

Customization tip: Add “I live in [LOCATION]” — ingredient availability varies significantly by region, and Claude will suggest locally available alternatives.

48. Travel Itinerary

Plan a [NUMBER]-day trip to [DESTINATION]. Travel dates: [DATES OR SEASON]. Budget: [$ PER DAY or TOTAL]. Travel style: [budget backpacker/mid-range comfort/luxury]. Interests: [LIST — e.g., food, history, nature, nightlife, art]. Constraints: [ANY — mobility issues, traveling with kids, dietary restrictions]. Create a day-by-day itinerary with: specific locations and attractions (not generic "explore the city"), recommended restaurants with cuisine type and price range, transportation between locations with estimated time and cost, one off-the-beaten-path recommendation per day (something most tourists miss), daily budget breakdown, and one "rest moment" per day (you can't sprint for 12 hours). Also include: 3 things to book in advance, 3 local customs or etiquette tips, and packing essentials specific to this destination and season.

What it does: Creates a realistic, budget-tracked travel itinerary with specific venues, local tips, and practical logistics.

Why it works with Claude: The “rest moment” instruction prevents the overloaded AI itinerary problem. Claude’s knowledge of geographic distances means it won’t schedule sites on opposite ends of a city back-to-back — a common failure in other AI travel planners.

Customization tip: After generating, ask: “Now create an alternative rainy-day plan for any outdoor-heavy days” for weather resilience.

49. Gift Ideas

Suggest 10 gift ideas for [DESCRIBE THE PERSON — relationship, age, interests, personality]. Budget: [RANGE]. Occasion: [BIRTHDAY/HOLIDAY/THANK YOU/etc.]. For each suggestion: Name the specific product or experience (not generic "a book" — which book?), explain why it fits this specific person (connect to their interests/personality), provide a price estimate, and rate the "thoughtfulness factor" — will this feel generic (3/10) or deeply personal (10/10)? Include a mix of: 3 physical gifts, 3 experience gifts, 2 consumable gifts (food, drink, subscription), and 2 wildcard/creative gifts they would never think of themselves. Avoid: [LIST ANYTHING TO EXCLUDE — e.g., "they already have too many books," "no alcohol"]. End with your #1 recommendation and why.

What it does: Generates specific, thoughtful gift recommendations with personalization reasoning and a diverse gift category mix.

Why it works with Claude: The “thoughtfulness factor” rating forces Claude to evaluate each suggestion against the specific person, not just the occasion. Claude’s wildcard suggestions are often genuinely creative because its training includes diverse cultural knowledge about gifting traditions worldwide.

Customization tip: Add their social media bio or a description of their home/office — Claude will pick up on personality cues and style preferences.

50. Meal Planning

Create a 7-day meal plan. Dietary requirements: [LIST]. Weekly grocery budget: [AMOUNT]. Cooking skill: [beginner/intermediate/advanced]. Time available: [WEEKNIGHT COOKING TIME — e.g., "30 minutes max on weekdays, more time on weekends"]. Household size: [NUMBER OF PEOPLE]. For each day, provide: breakfast, lunch, dinner, and one snack. Each meal should include: the dish name, approximate prep+cook time, calorie estimate, and the recipe (ingredient list + brief instructions). Design the plan so that: ingredients overlap between meals (minimize waste), prep-ahead opportunities are noted ("cook extra rice on Monday for Wednesday's stir-fry"), the most complex meals are on weekends, and the grocery list is consolidated and organized by store section. End with the total weekly grocery list with estimated costs.

What it does: Produces a complete weekly meal plan with recipes, grocery list, prep-ahead strategies, and waste-minimizing ingredient overlap.

Why it works with Claude: Claude’s ability to plan across all 7 days simultaneously means ingredients genuinely overlap between meals. The time-aware scheduling (complex meals on weekends) shows Claude’s capacity for practical constraint satisfaction across an entire planning horizon.

Customization tip: Add “We eat [CUISINE PREFERENCES] and dislike [SPECIFIC FOODS]” for a plan you’ll actually enjoy following. After the first week, ask Claude to generate Week 2 that introduces variety while reusing any leftover ingredients.

---

How to Get the Most From These Prompts

These 50 prompts are starting templates, not rigid scripts. To get better results from Claude, apply these prompt engineering principles from our AI prompt library:

• Fill in every bracket. The more specific your [CONTEXT], the more specific Claude’s output. “Write about marketing” produces generic content. “Write about email marketing for B2B SaaS companies with under $1M ARR targeting HR directors” produces actionable content.
• Paste, don’t summarize. Claude’s 200K context window is there for a reason. Paste the full document, the complete email chain, the entire codebase. Context is Claude’s superpower.
• Iterate in the same conversation. Claude remembers context within a conversation. After the first output, say “This is good, but make it more casual” or “Expand section 3 and cut section 5.” You are collaborating, not ordering from a menu.
• Tell Claude what you DON’T want. Negative constraints (“don’t use jargon,” “don’t start with a question,” “no bullet points in this section”) are powerful with Claude. It respects exclusions better than most models.
• Request your preferred format up front. If you need a table, say “format as a table.” If you need markdown, say “use markdown headers.” If you need code blocks, specify the language. Claude will match your format precisely.

---

Download the Complete Claude Essentials Guide

Want all 50 prompts in a printable PDF, plus bonus templates for chaining prompts together? Our 50 AI Prompts collection includes these prompts and more, formatted for quick reference.

---

Frequently Asked Questions

What makes a good Claude prompt?

A good Claude prompt has four elements: clear context (who you are, what you need), specific instructions (not vague direction), defined output format (how you want the response structured), and quality constraints (what to include and what to avoid). Claude’s Constitutional AI training means it follows multi-part instructions more faithfully than most models, so the more specific you are, the better your results. Research from Anthropic shows that prompts with explicit structure requirements produce outputs that are 40% more likely to match user expectations on the first attempt. The key difference between a good and great Claude prompt is specificity — replace every vague word (“good,” “detailed,” “professional”) with a concrete benchmark (“above 90% test coverage,” “under 200 words,” “in the tone of a McKinsey consultant”).

Are Claude prompts different from ChatGPT prompts?

Yes, in meaningful ways. Claude and ChatGPT are trained using different methodologies — Claude uses Constitutional AI while ChatGPT uses RLHF with human labelers — which creates different prompting sweet spots. Claude handles longer prompts and more complex instructions better, so you can give Claude a detailed multi-step prompt that would confuse other models. Claude also responds better to negative constraints (“don’t do X”) and role-based instructions (“act as a senior engineer reviewing code”). ChatGPT tends to produce more enthusiastic, marketing-style language by default, while Claude produces more measured, analytical language. Neither is better — they are different tools. For a detailed comparison, see our ChatGPT vs Claude vs Gemini comparison. If you are migrating prompts from ChatGPT to Claude, the main change is that you can provide significantly more context and ask for more structured outputs.

How long should a Claude prompt be?

As long as it needs to be — with Claude, longer prompts typically produce better results, up to a point. Claude’s 200K-token context window (roughly 150,000 words) means you can provide extensive context without truncation. For most tasks, a prompt between 100 and 500 words hits the sweet spot: enough detail to eliminate ambiguity but not so much that critical instructions get buried. For complex tasks like code reviews or document analysis, paste the full source material and keep your instructions concentrated at the beginning or end of the prompt (Claude pays strongest attention to the beginning and end of long contexts, a pattern noted by Stanford research on long-context LLMs, arXiv:2307.03172). A one-sentence prompt almost always produces a generic response. If your prompt is under 50 words, you are probably leaving quality on the table.

Can I save prompts in Claude?

As of March 2026, Claude on claude.ai does not have a built-in prompt library or saved templates feature. However, there are effective workarounds: you can use Claude’s “Projects” feature (available on Pro and Team plans) to save system prompts that apply to every conversation within a project — this effectively saves your most-used prompts as persistent context. For individual prompts, users typically save them in a note-taking app (Notion, Apple Notes, Google Docs) or a dedicated prompt management tool. The Claude API supports “system prompts” that can be saved programmatically. Anthropic has hinted at expanded personalization features on their product roadmap. For a curated, ready-to-use prompt collection, check our AI prompt library which you can bookmark and reference anytime.

What is Claude’s system prompt?

Claude’s system prompt is a set of instructions that Anthropic loads before every conversation to define Claude’s baseline behavior — its personality, safety guidelines, and default response patterns. You cannot see or modify Anthropic’s system prompt on claude.ai, but you can effectively override parts of it by providing your own instructions at the start of a conversation or using the Projects feature. When using the Claude API, developers can write custom system prompts that set the tone, role, and boundaries for Claude’s responses in their application. According to Grokipedia’s entry on large language models, system prompts are the primary mechanism through which AI providers shape model behavior post-training. Understanding that system prompts exist helps explain why Claude behaves differently on claude.ai versus in third-party apps that use the Claude API — they are using different system prompts.

---

Start Using Better Prompts Today

The difference between a mediocre AI experience and a great one is prompt quality. These 50 prompts give you a head start, but the real skill is learning to write your own. Pay attention to which prompts produce the best results for your specific needs, modify them, and build your personal prompt library over time. Claude rewards specificity, context, and structure — give it those three things, and it will consistently deliver outputs worth using.

---

More Claude Guides from Beginners in AI

• How to Use Claude AI: Complete Guide
• Claude AI Login: How to Sign Up and Get Started
• Claude App: iOS, Android, and Desktop Guide
• Claude API for Beginners
• Claude Code: Beginner’s Guide
• ChatGPT vs Claude vs Gemini