CLAUDE PROMPTS

Become a better
writer
with Claude

The AI That Thinks Before It Writes

Claude leads benchmarks for long-document comprehension
Anthropic 2025
Professionals spend 60% less time editing Claude output
Independent user studies 2025
80% of Fortune 500 use Claude for legal and strategic drafts
Anthropic 2025

What You Can Do with Claude

Real tasks. Real prompts. Copy, paste, and get results in minutes.

1

Write a complete PRD from a 15-word feature idea

How to use this

  1. Open Claude (claude.ai). Start a new conversation.
  2. Copy the prompt below. Replace the feature idea in brackets with your own — keep it to one or two sentences.
  3. Claude will output a structured PRD with goals, user stories, acceptance criteria, edge cases, and out-of-scope items.
  4. Share directly with engineers or paste into your project management tool (Linear, Jira, Notion). Ask Claude follow-up questions to expand any section.
PRD Generator
You are a senior product manager at a B2B SaaS company with 10 years of experience writing PRDs that engineers love to build from.

FEATURE IDEA:
[YOUR FEATURE IN 1–2 SENTENCES — e.g. "Allow users to set a recurring 'focus block' in the app that automatically blocks notifications and marks them as unavailable in shared calendars"]

MY CONTEXT:
- Product: [YOUR PRODUCT NAME + one sentence description]
- Target user: [WHO USES THIS FEATURE — e.g. "Individual contributors who manage their own calendar"]
- Current workaround: [HOW THEY SOLVE THIS TODAY — e.g. "Manually setting Do Not Disturb and blocking calendar time separately"]
- Success metric: [HOW WE'LL KNOW IT WORKED — e.g. "30% of active users set at least one focus block within 7 days of feature launch"]

Write a complete PRD with the following sections:

## 1. OVERVIEW
- Problem statement (the pain in the user's words, not product-speak)
- Proposed solution (one paragraph)
- Why now (what makes this the right time to build this)

## 2. GOALS & NON-GOALS
- 3–5 goals (what this feature must accomplish)
- 3–5 explicit non-goals (what we are NOT building in this version)
- Success metrics (measurable outcomes, not vanity metrics)

## 3. USER STORIES
Format each as: "As a [user type], I want to [action] so that [outcome]."
- 3–5 primary user stories (must-have scenarios)
- 2–3 edge case user stories (error states, unusual inputs, permission scenarios)

## 4. FUNCTIONAL REQUIREMENTS
- List every specific behavior the feature must have
- Number each requirement
- Mark each as P0 (must ship), P1 (should ship), or P2 (nice to have)

## 5. ACCEPTANCE CRITERIA
For each P0 requirement, write specific, testable acceptance criteria.
Format: "Given [context], when [action], then [expected result]."

## 6. OPEN QUESTIONS
- List 5–7 questions that need answers before or during development
- Flag who owns each question (PM, Engineering, Design, Legal)

## 7. OUT OF SCOPE
- Explicitly list what this PRD does NOT cover
- Note which out-of-scope items belong in a future version

Keep language precise and unambiguous. Engineers should be able to start building from this document without a single clarifying meeting.

Why this works: One paragraph of context, a complete PRD out. The user story format, acceptance criteria structure, and explicit non-goals are what make this document buildable — not just readable. Engineers get exactly what they need to start without a kickoff meeting. A senior PM would spend 3–4 hours writing this; Claude returns it in under a minute.

2

Rewrite any AI-sounding draft to pass the human test

ClaudeClaudeChatGPTChatGPT·Course 57 — Stop Sounding Like AI

How to use this

  1. Open Claude (claude.ai) or ChatGPT (chatgpt.com). Start a new chat.
  2. Copy the prompt below. Paste your draft in the DRAFT TO REWRITE section.
  3. Fill in the VOICE PROFILE section — the more specific you are about tone and banned words, the better the output.
  4. Review the rewritten version. Ask follow-up: 'Make it 20% more direct' or 'Replace the opening — it still sounds generic' to iterate.
De-AI Rewrite
You are an expert editor who specializes in making AI-generated text sound genuinely human.

DRAFT TO REWRITE:
[PASTE YOUR DRAFT HERE]

VOICE PROFILE (fill in):
- Tone: [e.g. "Direct and slightly irreverent — like a smart friend who's been in the industry 10 years"]
- Audience: [e.g. "Senior marketers at mid-size B2B companies"]
- Never say: [YOUR BANNED PHRASES — e.g. "leverage," "utilize," "dive deep," "game-changer," "in today's fast-paced world"]
- Always include: [YOUR SIGNATURE STYLE — e.g. "Short sentences. Specific numbers. End with a question or a clear CTA, never a summary."]
- Writing sample for reference: [PASTE 2–3 SENTENCES YOU'VE WRITTEN THAT SOUND LIKE YOU — or leave blank]

REWRITE INSTRUCTIONS:

1. STRIP THESE PATTERNS FIRST:
- Filler openers: "In today's world...", "It's no secret that...", "As we know..."
- False transitions: "Moreover," "Furthermore," "In conclusion,"
- Hedge phrases: "It's important to note," "It's worth mentioning," "One could argue"
- Generic adjectives: "innovative," "cutting-edge," "seamless," "robust," "best-in-class"
- Redundant pairs: "each and every," "first and foremost," "tried and true"

2. THEN APPLY THESE RULES:
- Replace abstract nouns with specific verbs (not "optimization" but "cut from 40 steps to 12")
- Replace vague quantities with real numbers or ranges
- Replace passive voice with active voice
- Make every sentence earn its place — if deleting it doesn't change the meaning, delete it
- Keep the first sentence under 15 words
- The last sentence should land, not trail off

3. OUTPUT FORMAT:
- Rewritten version (full text)
- 3 sentences you changed the most and why
- Any lines you flagged as still-risky (might still sound AI-generated)

Unlock this prompt

No spam. Just the prompt.

Why this works: Most AI rewrites just swap words. This prompt goes deeper: it strips the structural patterns that signal AI (hedge phrases, false transitions, abstract nouns) before rebuilding in your specific voice. The output section forces the AI to flag its own weakest lines — so you know exactly where to focus your editing pass.

3

Extract key decisions and actions from any long document

How to use this

  1. Open Claude (claude.ai). Claude handles long documents better than any other model — up to 200K tokens of context.
  2. Copy the prompt below, then paste your document (meeting transcript, report, contract, research paper) after the prompt text.
  3. Claude will extract decisions, action items, open questions, and key data points in a structured format.
  4. Export the output directly to Notion, Google Docs, or your project tool. Send it to your team — no meeting recap needed.
Document Intelligence Extractor
You are a senior business analyst. I'm going to paste a long document below this prompt. Your job is to extract everything actionable from it in a structured format.

DOCUMENT TYPE: [WHAT THIS IS — e.g. "Board meeting transcript," "Quarterly business review," "Due diligence report," "Research paper," "Legal contract"]

MY ROLE: [YOUR ROLE — e.g. "I'm the Head of Product and need to extract my team's action items and any product decisions made"]

WHAT I CARE MOST ABOUT: [YOUR PRIORITY — e.g. "Decisions that affect our Q2 roadmap and any commitments made to the board"]

After reading the document, extract and structure the following:

## 1. DECISIONS MADE
For each decision:
- Decision: what was decided (one sentence)
- Made by: who made it (if named)
- Rationale: why (if stated)
- Effective date: when it takes effect (if mentioned)

## 2. ACTION ITEMS
For each action item:
- Owner: who is responsible
- Task: what they need to do (specific and actionable)
- Deadline: by when (exact date if mentioned, "ASAP" or "next meeting" if implied)
- Dependencies: what needs to happen first (if any)
- Priority: HIGH / MEDIUM / LOW based on context

## 3. OPEN QUESTIONS
- Questions that were raised but not resolved
- Who owns finding the answer
- Why it matters

## 4. KEY DATA POINTS
- Important numbers, metrics, or statistics mentioned
- Context for each (what does this number mean in this situation?)

## 5. COMMITMENTS MADE
- Any promises, guarantees, or commitments made by named parties
- To whom and by when

## 6. RISKS FLAGGED
- Any risks, concerns, or red flags mentioned
- Who raised them and how seriously they were treated

## 7. EXECUTIVE SUMMARY
- 5–7 bullet points capturing the most important things in the entire document
- Write this for someone who won't read anything else

If anything is ambiguous, note the ambiguity. Do not invent details that aren't in the document.

--- DOCUMENT STARTS BELOW ---
[PASTE YOUR DOCUMENT HERE]

Available when AI School launches

This prompt will be available in the app. Join the waitlist for early access.

Join the Waitlist

Why this works: Claude's 200K context window means it can read what other models can't — full meeting transcripts, lengthy contracts, complete research reports. This prompt extracts every decision, action, commitment, and risk with enough structure that the output goes directly into your project tool. The final executive summary means your team gets the full picture in 30 seconds.

4

Run a full competitive analysis in 30 minutes

ClaudeClaudePerplexityPerplexity·Course 19 — Real-Time Competitive Intel

How to use this

  1. Run PROMPT 1 in Perplexity (perplexity.ai) to gather live data on each competitor — pricing, features, positioning, recent news.
  2. Copy the full Perplexity output. Open Claude (claude.ai) and paste PROMPT 2 with the research included.
  3. Claude will synthesize the raw data into a structured analysis: positioning gaps, messaging weaknesses, and your 3 best attack vectors.
  4. Share the output with your team as your competitive brief — review and update monthly.
Competitive Intelligence Sprint
--- PROMPT 1: Research (run in Perplexity) ---

Research these competitors for [YOUR COMPANY — e.g. "a B2B project management tool for construction companies"]:

Competitors: [LIST YOUR TOP 3–5 — e.g. "Procore, Buildertrend, CoConstruct, PlanGrid"]

For each competitor, find:
1. Current pricing tiers and what each includes
2. Primary messaging and positioning (headline + tagline from their homepage)
3. Top 3 features they lead with
4. Target customer (who they say they're for)
5. Recent news: funding, product launches, partnerships, or layoffs in the past 6 months
6. Top customer complaints (check G2, Capterra, or Trustpilot reviews — quote real reviews)
7. Any known weaknesses or controversies

Cite sources for every data point.

--- PROMPT 2: Analysis (run in Claude — paste PROMPT 1 output below) ---

You are a competitive intelligence analyst. Here is research on our competitors:

[PASTE FULL PERPLEXITY OUTPUT HERE]

MY COMPANY:
- What we do: [ONE SENTENCE]
- Our target customer: [SPECIFIC — job title, company size, situation]
- Our key differentiator: [WHAT WE DO THAT THEY DON'T]
- Our current positioning: [OUR HOMEPAGE HEADLINE]

Analyze this data and produce:

## 1. COMPETITIVE LANDSCAPE MAP
- How each competitor positions itself (their narrative in one sentence)
- Where they cluster (what messaging themes repeat across multiple competitors?)
- The "white space" — what positioning no one owns

## 2. THEIR WEAKNESSES (ranked by exploitability)
- For each competitor: top 2 weaknesses based on real customer complaints
- Which weaknesses are structural (hard to fix) vs. tactical (easy to fix)

## 3. OUR ATTACK VECTORS
- Top 3 specific ways we can differentiate in messaging and product
- For each: what to say, who to say it to, which channel to use

## 4. THREATS TO US
- What are they doing that could hurt us in the next 6–12 months?
- Which threat is most serious and why?

## 5. RECOMMENDED POSITIONING MOVE
- One specific change to our positioning or messaging that would sharpen our differentiation
- The exact headline rewrite you'd recommend testing

Available when AI School launches

This prompt will be available in the app. Join the waitlist for early access.

Join the Waitlist

Why this works: Perplexity handles live research with citations; Claude handles synthesis. Using them in sequence means your competitive analysis is built on real current data (not hallucinated facts) and interpreted by a model that can reason about positioning strategy. The two-prompt structure alone saves 3–4 hours vs. doing this manually.

5

Generate a complete content strategy from one brand brief

How to use this

  1. Open Claude (claude.ai). Start a new conversation.
  2. Copy the prompt below. Fill in the BRAND BRIEF section with your product and audience details.
  3. Claude will output a 30-day content strategy: themes, post types, platform-specific guidelines, and a week-by-week calendar.
  4. Export to Notion or Google Sheets. Share with your content team or use it solo with the Content Spine prompt to generate each week's posts.
Content Strategy Builder
You are a content strategist with deep experience in B2B and B2C brand building. I'm going to give you a brand brief and I need a complete 30-day content strategy.

BRAND BRIEF (fill in all fields):
- Brand name: [YOUR BRAND]
- What it does: [ONE SENTENCE — what problem you solve]
- Target audience: [SPECIFIC — job title, age range, situation, pain point]
- Brand voice: [3 adjectives that describe how you sound — e.g. "Direct, empathetic, data-driven"]
- Content goal: [WHAT SUCCESS LOOKS LIKE — e.g. "Build an audience of 5K followers who see us as the go-to source for remote work productivity"]
- Primary platforms: [WHERE YOU'RE POSTING — e.g. "LinkedIn (primary), Twitter/X (secondary), Email newsletter (weekly)"]
- What you will NOT do: [CONTENT BOUNDARIES — e.g. "No political content, no personal life sharing, nothing that requires video production"]

Build a COMPLETE 30-DAY CONTENT STRATEGY:

## 1. CONTENT PILLARS (4–5 pillars)
For each pillar:
- Name and description
- Why it serves the audience
- What formats work best for this pillar
- Percentage of total content volume

## 2. PLATFORM STRATEGY
For each platform listed:
- Optimal posting frequency
- Best content types for that platform
- Tone adjustments (same voice, different register)
- Primary metric to track

## 3. CONTENT CALENDAR (4 weeks)
For each week:
- Theme (what unifying idea connects this week's content?)
- Post schedule: day, platform, pillar, format, and one-sentence topic
- One "anchor piece" (the highest-effort post of the week that everything else supports)

## 4. CONTENT FORMATS TO USE
- For each format: description, effort level (low/medium/high), expected engagement type
- Recommended mix for this brand's goals

## 5. THE FIRST 5 POSTS (write them in full)
- These should be immediately publishable
- Each from a different pillar
- Include: post copy, hashtags, and any visual direction

## 6. WHAT TO MEASURE
- 3 metrics that signal the strategy is working
- 3 early warning signs that something needs to change
- When to review and adjust (suggested cadence)

Available when AI School launches

This prompt will be available in the app. Join the waitlist for early access.

Join the Waitlist

Why this works: Most content strategies stay at the level of 'post 3x per week on LinkedIn.' This one goes all the way to a day-by-day calendar with pillars, formats, platform-specific rules, and five fully written posts you can publish today. Claude's long context window means it holds all the brand context across every section — so the calendar actually matches the strategy.

Featured Courses

Master Claude — start with these.

Get notified when AI School launches.

Be the first to access 66+ hands-on AI courses.

Join the Waitlist
66+ courses50+ AI tools15 min avg session

Frequently Asked Questions

What is Claude best used for compared to ChatGPT?

Claude excels at long-document tasks (contracts, transcripts, research papers), technical writing that requires precision (PRDs, specs, briefs), and producing copy that sounds genuinely human rather than AI-generated. ChatGPT is faster for rapid iteration and stronger for structured data tasks. For anything over 10,000 words or anything where 'human-sounding' matters, Claude is the better choice.

How do I write good prompts for Claude?

Claude responds best to prompts that include: a specific role ('you are a senior product manager'), rich context about your situation (company stage, audience, goal), a clear output structure (sections, formats, word counts), and explicit constraints (what to avoid, what to always include). Unlike ChatGPT, Claude handles very long context windows well — so don't be afraid to paste in full documents alongside your instructions.

Can Claude write a PRD or technical specification?

Yes — Claude is one of the strongest models for PRD writing. Give it the feature idea, user context, success metrics, and your engineering team's preferences, and it will produce user stories, acceptance criteria, edge cases, and explicit non-goals. The output is typically buildable without a kickoff meeting, which is the real test of a good PRD.

Is Claude free to use?

Claude has a free tier (Claude.ai) with access to Claude 3 Sonnet with daily usage limits. Claude Pro ($20/month) gives you access to Claude 3.5 Sonnet and Opus, priority access during peak hours, and significantly higher usage limits. For professional use — especially long documents and complex reasoning tasks — the Pro tier is worth it.

How do I make AI-generated content sound more human with Claude?

The key is a two-step process: first, identify and strip the structural patterns that signal AI (hedge phrases like 'it's important to note,' false transitions like 'furthermore,' and vague adjectives like 'robust'). Second, rebuild with specific verbs instead of abstract nouns, real numbers instead of vague quantities, and active voice throughout. Claude's De-AI rewrite prompt handles both steps automatically when you feed it your voice profile.