Why 90% of people get terrible AI results (and how to fix it)
I've written over 2,000 AI prompts this year. Here's what I've learned:
Most people type something vague like "write me an email" and then complain that AI is useless. The problem isn't the AI — it's the instruction.
The 3 things that separate a garbage prompt from a pro-level one:
Role assignment — Tell the AI WHO it is. "You are a senior copywriter with 15 years of experience in SaaS" gets you wildly different results than just asking for copy.
Context loading — Give it the background. Your audience, your tone, your constraints. The more context, the less generic the output.
Output formatting — Specify exactly what you want back. Bullet points? A table? A 3-paragraph email? Don't leave it guessing.
Here's a quick example:
❌ "Write a LinkedIn post about productivity"
✅ "You are a thought leader in remote work productivity. Write a LinkedIn post (150 words max) sharing one counterintuitive productivity tip. Use a conversational tone, start with a bold claim, and end with a question to drive comments."
The difference in output quality is night and day.
I've been building a library of prompts that follow these principles — tested across ChatGPT, Claude, Midjourney, and more. If you're tired of mediocre AI outputs, The Prompt Lab has everything you need to actually get results.
