The Structure of a Great Prompt
Most AI outputs sound generic because the prompt was generic. Here's the structure that actually works.
Context files
Upload anything that defines you or the task:
- About you (background, voice, what you do)
- Anti-AI writing style (rules of what NOT to do)
- Copywriting principles (frameworks you trust)
The model can't sound like you if it doesn't know who you are.
The task in one clear sentence
"I want to write a LinkedIn post about AI tools I use daily." Specific topic, specific platform, specific angle. Not "write me content."
Order of operations
Tell the model what to do FIRST before generating anything:
- Read all uploaded files completely
- Don't start executing yet
- Ask clarifying questions before writing
This single rule kills 90% of bad outputs.
The clarification loop
Use the AskUserQuestion approach so the model asks one question at a time. You answer, it refines, it asks the next. By the time it writes, the angle is already aligned with you.
Generation
Only after context, task, rules, and clarification — let it write. The first draft will be 10x closer to what you actually wanted.
Pro tip: save your context files in a Project. Every new chat in that Project automatically has them. You write the prompt once and reuse forever.
