How to Write Prompts That Get 10x Better Results
The exact 5-part framework we use to test every prompt in our packs.
1. Assign a Role
Start every prompt by telling the model who to be. "You are a senior conversion copywriter with 10 years of experience in SaaS landing pages" outperforms a bare request because it narrows the model's behavior to a proven style. The role sets tone, vocabulary and depth before a single task is given.
2. Give Context
The model can't read your mind. Include the audience, the product, the goal and any constraints. "Write a headline for a project-management tool aimed at freelance designers who hate meetings" gives the model four anchors to work from. Vague context produces vague output — every time.
3. State the Task Precisely
One task per prompt. "Write a 300-word product description with a benefit-led opening, three feature bullets, and a closing CTA" is a task. "Describe my product" is a wish. Specificity is the single biggest quality lever in prompt writing.
4. Set Constraints
Tell the model what to avoid: word count, tone to skip, formats to exclude, claims not to make. Constraints prevent the two most common failure modes — rambling and hallucinated specifics. "No jargon, no exclamation marks, under 120 words" is a constraint block that works.
5. Define the Output Format
Ask for a table, a list, JSON, or a specific structure and the model will comply. "Return the answer as a markdown table with columns: Task, Model, Winner, Why" turns a wall of text into something you can actually use. Format requests are the cheapest quality boost available.
Want the Prompts, Not the Theory?
Every technique in this article is baked into our prompt packs — tested, organized, and ready to copy.
Browse Prompt Packs