How to Write Better AI Prompts (10 Tips That Actually Work)

Why Most People Get Mediocre AI Results

The difference between a useless AI response and one that saves you an hour comes down to how you ask. Most people type vague requests into ChatGPT or Claude and then complain that AI is overhyped. The AI isn't the problem -- the prompt is. Here are 10 techniques that consistently produce better results, with specific examples you can steal.

1. Give Context Before the Task

Don't jump straight to the ask. Set the scene first.

Weak: "Write a product description for headphones."

Strong: "I sell premium wireless headphones to audiophiles aged 25-45. Our brand voice is technical but approachable. Write a 100-word product description for our new noise-cancelling model that emphasizes sound quality and battery life."

The more relevant context the AI has, the less it has to guess. Less guessing means fewer generic responses.

2. Specify the Format You Want

AI will default to whatever format seems most common. If you want something specific, say so explicitly.

Examples: "Give me a numbered list of 5 items," "Format this as a comparison table with columns for Name, Price, and Best Feature," "Write in short paragraphs of 2-3 sentences max," "Use bullet points, not prose."

3. Define the Tone and Audience

The same information sounds completely different for a CEO versus a college student. Tell the AI who will read the output.

Try: "Explain this like I'm a marketing manager who's never used an API before" or "Write this for a technical blog audience that already knows Python basics."

4. Show an Example of What You Want

This is the single most underused prompting technique. Paste an example of the style, structure, or output you like and say: "Follow this format and tone, but write about [your topic]." Both ChatGPT and Claude are excellent at pattern-matching from examples.

5. Set Constraints

Constraints force creativity and prevent rambling. Give the AI guardrails.

Useful constraints: "Keep it under 200 words," "Use no more than 3 bullet points," "Don't use jargon," "Avoid cliches like 'game-changer' and 'revolutionize'," "Write at an 8th-grade reading level."

6. Ask for Multiple Options

Instead of asking for one answer, ask for three. "Give me 3 different subject lines for this email, ranging from formal to casual." This gives you choices and often surfaces ideas you wouldn't have considered.

7. Use the Chain-of-Thought Approach

For complex problems, ask the AI to think step by step before giving its final answer. "Walk me through your reasoning before providing a recommendation" produces dramatically better analysis than "What should I do?" This works especially well for strategy questions, debugging, and data analysis.

8. Iterate Instead of Regenerating

When the first response isn't quite right, don't start over. Build on it. "Good structure, but make the opening more compelling and add a specific example in section 2" is far more effective than hitting regenerate and hoping for the best. Treat the conversation as a collaboration, not a slot machine.

9. Assign a Role

Telling the AI to adopt a specific perspective changes the depth and angle of its response.

Examples: "You are a senior data analyst reviewing this spreadsheet," "Act as a skeptical investor evaluating this business plan," "You are a UX researcher conducting a heuristic evaluation of this interface."

This technique works particularly well in Claude, which tends to stay in character consistently throughout long conversations.

10. Use the Meta-Prompt

When you're not sure how to prompt for what you need, ask the AI to help you build the prompt. Try: "I want to [describe your goal]. What information do you need from me to give the best possible result?" The AI will ask clarifying questions, and your answers become the perfect prompt. This is especially powerful in Perplexity for research tasks and in ChatGPT for creative work.

A Quick Prompt Formula

If you remember nothing else, use this structure for every prompt:

• Context: Who you are, what the project is
• Task: Exactly what you want the AI to do
• Format: How you want the output structured
• Constraints: Length, tone, what to avoid

That four-part structure will get you better results than 90% of users, regardless of which AI tool you're using. For a deeper comparison of the best tools to prompt, see our ChatGPT vs Claude comparison.