How to Write Better AI Prompts

The core problem with vague prompts

"Write me a marketing email" gives the model almost nothing to work with — no audience, no product, no tone, no constraint on length. It will guess, and the result will read as generic because it is generic. Specificity is the single highest-leverage change you can make to prompt quality.

A structure that works for most tasks

1. Role or context (optional but often helpful): "You're helping me write..."
2. The task, stated precisely: what output do you want, exactly?
3. Constraints: length, tone, audience, format, things to avoid.
4. Relevant input: the actual content to work from (a draft, notes, data).
5. Output format: bullet points? A table? Plain prose?

Example: before and after

Vague: "Write a product description for my headphones."

Specific: "Write a 60-word product description for wireless noise-cancelling headphones, targeting commuters who work remotely part-time. Emphasize battery life and comfort for long calls. Avoid superlatives like 'best' or 'revolutionary' — focus on concrete, specific benefits."

The second version gives the model an actual brief instead of a title.

Iterate instead of restarting

If the first response isn't right, it's usually faster to give feedback ("make this more concise" / "this doesn't emphasize the battery life enough") than to write an entirely new prompt from scratch. Treat the conversation as a working session, not a single query-response pair.

Use variables for prompts you'll reuse

If you find yourself writing a similar prompt repeatedly, turn the changing parts into explicit variables (like {{topic}} or {{audience}}) and save the template. Our prompt library is organized exactly this way — every prompt documents its variables so you can adapt it quickly.