How to Write AI Prompts That Actually Work

Most people get mediocre results from AI because they write mediocre prompts.

I’ve written thousands of prompts. Here’s what actually works.

The Core Problem

Bad prompt:

Write about productivity

Result: Generic, unhelpful, could-be-written-by-anyone content.

Good prompt:

Write a 600-word blog post about time blocking for software developers who work from home. Include a morning routine example, common mistakes, and how to handle interruptions. Tone: practical and direct, like advice from a senior colleague.

Result: Specific, useful, targeted content.

The difference is specificity.

The Prompt Framework

Every good prompt has these elements:

1. Task (Required)

What do you want the AI to do?

• “Write a blog post about…”
• “Summarize this document…”
• “Create a list of…”
• “Analyze this data and…”
• “Draft an email to…”

Be specific about the action.

2. Context (Usually helpful)

What background does the AI need?

• Who is the audience?
• What’s the purpose?
• What’s the situation?

Example: “I’m a freelance designer emailing a client who hasn’t paid their invoice in 30 days.”

3. Format (Often important)

How should the output be structured?

• Length (word count, number of items)
• Structure (bullet points, paragraphs, sections)
• Style (formal, casual, technical)

Example: “500 words, with 3 main sections and bullet points for key takeaways”

4. Tone (When it matters)

How should it sound?

• Professional but friendly
• Technical and precise
• Conversational and casual
• Authoritative and confident

Example: “Write like you’re explaining to a smart friend, not lecturing”

5. Constraints (Prevent problems)

What should the AI avoid?

• “Don’t use jargon”
• “Don’t include generic advice”
• “Don’t start with ‘In today’s world’”
• “Must include specific examples”

Real Prompt Examples

For Blog Posts

Weak:

Write about email marketing

Strong:

Write an 800-word blog post titled “5 Email Subject Lines That Actually Get Opened”

Audience: Small business owners who send their own marketing emails

For each subject line type:

• Explain why it works
• Give 2 specific examples
• Include one common mistake to avoid

Tone: Direct and practical. No fluff. Write like a marketing consultant giving honest advice.

Don’t use: “In today’s digital landscape”, generic statistics, or obvious advice like “keep it short”

For Emails

Weak:

Write an email to a client

Strong:

Write a follow-up email to a client who received a proposal 5 days ago and hasn’t responded.

Context: I’m a web designer. The proposal was for a website redesign at $8,000. The client seemed interested in our call but mentioned budget concerns.

Goal: Get them to either accept, negotiate, or tell me they’re not interested. Don’t be pushy.

Length: Under 100 words Tone: Professional but warm

For Analysis

Weak:

Analyze this data

Strong:

Analyze this sales data from Q1 2026 [paste data]

I need:

1. Top 3 trends you notice
2. Any concerning patterns
3. Recommended actions based on the data

Format: Bullet points, brief explanations Audience: I’m presenting this to our CEO who doesn’t like technical jargon

For Code

Weak:

Write a function to process users

Strong:

Write a Python function that:

• Takes a list of user dictionaries
• Filters to users who signed up in the last 30 days
• Returns their email addresses

Input format: [{“email”: “…”, “signup_date”: “YYYY-MM-DD”, “name”: “…”}]

Include error handling for missing fields Add docstring explaining the function Use type hints

Advanced Techniques

Chain of Thought

For complex reasoning, ask AI to show its work:

Analyze this business decision. Think through it step by step:

1. What are the potential benefits?
2. What are the risks?
3. What information is missing?
4. What’s your recommendation and why?

Role Assignment

Sometimes giving AI a role helps:

You’re a senior product manager at a SaaS company. Review this feature spec and identify any gaps or concerns a PM would typically catch.

Few-Shot Examples

Show AI what you want by example:

Convert these feature descriptions to user stories:

Example: Feature: “Password reset via email” User story: “As a user who forgot my password, I want to receive a reset link via email so I can regain access to my account”

Now convert: Feature: “Two-factor authentication” Feature: “Profile photo upload”

Iterative Refinement

Don’t expect perfection on try one. Follow up:

• “Make this more concise”
• “Add a specific example in paragraph 2”
• “The tone is too formal - make it conversational”
• “Expand on point 3”

Common Mistakes

1. Being too vague

Bad: “Help me with marketing” Good: “Create 5 Instagram captions for a coffee shop promoting their new cold brew”

2. Not specifying format

Bad: “Explain machine learning” Good: “Explain machine learning in 3 paragraphs suitable for a business executive with no technical background”

3. Forgetting context

Bad: “Write a thank you email” Good: “Write a thank you email to a hiring manager after a job interview for a marketing position. Mention that we discussed their upcoming product launch.”

4. Not setting constraints

Bad: “Write about productivity tips” Good: “Write about productivity tips. Don’t include obvious advice like ‘get enough sleep’ or ‘make a to-do list’”

5. One-and-done mentality

AI conversations are iterative. Your first prompt gets you started. Follow-ups get you to great.

Quick Reference

Basic structure:

[Task]: What to do
[Context]: Background information
[Format]: Length, structure
[Tone]: Voice, style
[Constraints]: What to avoid

Magic words that help:

• “Be specific”
• “Give examples”
• “Explain your reasoning”
• “Format as…”
• “Don’t include…”
• “Write as if…”

Bottom Line

Good prompts are specific prompts. Tell AI:

• Exactly what you want
• Who it’s for
• How it should be formatted
• What to avoid

The 2 minutes you spend writing a detailed prompt saves 20 minutes of editing mediocre output.

Start with the framework. Practice on real tasks. Your prompts will improve quickly.