Define the shape of the response
Specify whether you need a checklist, table, draft, script, comparison, or sequence of steps. Add a sensible length and request headings when the answer needs to be scanned quickly.
Use examples as direction, not decoration
A short example can communicate tone and quality faster than several adjectives. Explain what makes the example useful so the AI can apply the principle rather than copy its surface style.
Review, test, and refine
Treat the first answer as evidence. Identify what was missing, adjust one instruction at a time, and keep the prompt version that produces a reliable result. Always verify factual claims before publishing or acting.
Start with the result, not the request
Weak prompts often name an activity but not a destination. Instead of asking an AI to “write about customer research,” describe the decision the work should support. A strong objective gives the model a way to prioritize information.
Supply context that changes the answer
Useful context includes the audience, situation, available evidence, constraints, and what has already been tried. Include details only when they affect the result. More words do not automatically create a better prompt.
A simple prompt checklist
- Is the desired outcome explicit?
- Does the context change how the answer should be written?
- Are the audience and constraints named?
- Is the output format easy to follow?
- Will a human review facts and judgment calls?