Most people prompt ChatGPT the way they would search on Google.
It works, but the answers can be underwhelming.

The real shift happens when you change how you frame the question.
After recently reading Lee Bonnstra's prompt engineering guide for Google, I've made this quick cheat sheet for everyday users:
• 5 prompt types
• When to use each
• How to get better answers
• Examples for each
Here's a quick summary:
• Brainstorming = idea generation
• Role-based = specific tone/expertise
• Contextual = nuanced asks
• Step-by-step = logic-driven
• Instructional = action tasks
Any yes, combining them can get even better results.
Would love to hear your take - anything you would add?
Download to Lee Bonnstra's guide in the comments.
I’m Sam Page, CEO at 7DOTS - a digital innovation agency focused on building category-defining experiences.
I care deeply about how technology shapes the future - for business, for people, and for the planet.
If the post resonated, I’d love to hear your thoughts.
Or just drop me a message, I'm always happy to connect.
First published on LinkedIn.