Read before you buy.
Or skip straight to free.

Every professional-grade marketing prompt has five components. You don't need all five for every task. But when output quality matters, you want all five present.

1. Role

Tell Claude who it is. Not "you are a helpful assistant." That's the default and adds nothing. Assign a specific professional identity that shapes how Claude approaches the task.

• "You are a direct-response copywriter with 15 years of experience writing B2B SaaS landing pages."
• "You are a brand strategist who specializes in positioning challenger brands in crowded markets."
• "You are a performance marketing expert who writes email subject lines optimized for open rates."

Role assignment is not decoration. It activates different knowledge patterns and tones. A brand strategist and a direct-response copywriter approach the same brief in meaningfully different ways.

2. Context

Give Claude what it needs to know about your situation. Not background, operational intelligence. The more specific and relevant your context, the more specific and relevant your output.

Good context includes who the audience is (not "marketers," but "B2B marketing managers at companies with 50–500 employees who report to a VP or CMO"), the competitive landscape, what's already been tried, what constraints exist.

3. Constraints

Tell Claude what to avoid. Length. Tone. Banned words. Format requirements. Brand voice rules. The more rigid the constraint, the more useful the output usually is.

4. Examples

One good example beats ten paragraphs of description. Paste in a piece of work you love and say "use this voice." Claude is dramatically better at imitation than instruction.

5. Format

Specify the structure. Bullet points. Numbered list. Headers. A specific table. Word count. Without this, Claude defaults to whatever it thinks is helpful, which often isn't what you wanted.

Before you start using Claude for anything business-related, spend three minutes on these settings. They control how Anthropic handles your conversations.

Turn off conversation training (important)

By default, Anthropic may use your conversations to improve Claude's training. For a business owner, this means the content of your conversations, including client information, business plans, or anything else you type, could be reviewed by Anthropic staff or used in training data.

1. Click your profile or account menu
2. Go to Settings
3. Find Privacy or Data & Privacy
4. Find the option about conversation data being used for training
5. Toggle it OFF

If you handle sensitive client data

If you're on Claude Pro and you regularly handle sensitive client data, know that Anthropic offers API access with zero-retention configuration, meaning conversations are never stored. This requires technical setup. For most business owners, the standard privacy settings with training turned off are sufficient.

What never to paste. The full list.

• Client passwords, login credentials, or API keys of any kind
• Patient health information (PHI) on a non-business plan
• Unredacted financial records, account numbers, or tax IDs
• Trade secrets or strategic plans you wouldn't email a competitor
• Anything covered by an NDA you signed
• Source code that contains secrets, keys, or credentials embedded in it

Most marketing copy fails because it sounds like marketing copy. The fix is simple in theory and almost nobody does it: write in the exact words your audience uses to describe their own problems.

A Vocabulary File is a living document where you collect verbatim phrases your audience uses, from Reddit, reviews, calls, support tickets, podcasts. You build it once, then paste it into every prompt as context.

The 30-minute starter version

1. Pick a niche subreddit your audience hangs out in
2. Search for: "I hate when", "the worst part", "frustrating", "wish there was"
3. Copy the top 20 results into a doc. Exact words, no paraphrasing.
4. Tag each one: problem language, desired outcome, objection, moment of pain
5. Save. This is your starter Vocabulary File.

How you actually use it

Most non-coders using Cursor lose work at least once before they figure out the safety system. The system itself isn't complicated, but you have to set it up before you need it. After this, you'll never panic about losing changes again.

The three pieces

• Git (15 min): Your time machine. Every change saved as a checkpoint you can return to.
• .cursorrules (15 min): A file that tells Cursor how to behave in your project. Your style, your conventions, what to never touch.
• Knowledge files (30 min): Documents Cursor reads first every time. Project context, preferences, dependencies.

Day-one Git setup (no terminal heroics)

1. Install GitHub Desktop (free)
2. Connect your project folder
3. Make your first commit ("initial state")
4. Set up auto-commit on every Cursor save (instructions in the full guide)

Why this matters for non-coders specifically

When Cursor makes a change you didn't intend, and it will, especially early on, you need a way back that doesn't involve Ctrl+Z through 200 edits. Git is that way back. It costs nothing. Takes an afternoon to set up. Worth doing before your second session.