Role Prompting

The plain-English definition

Role prompting is opening your prompt with "Act as..." or "You are a..." to tell the AI what kind of expert or persona to embody. The model adapts its language, assumptions, and structure to match the role — pulling on patterns from how that role actually talks and thinks.

Why it works

An AI trained on the entire internet has read 10,000 cover letters written by hiring managers, 50,000 marketing emails written by copywriters, 20,000 legal contracts written by lawyers. When you set the role, you activate the patterns from that subset of training data. You're not giving the AI new knowledge — you're focusing it on the relevant parts of what it already knows.

Real example

No role: "Write a blog post about retirement planning."
Result: generic, hedged, encyclopedic.

With role: "You are a fee-only financial advisor with 20 years of experience working with middle-class clients. Write a blog post about retirement planning aimed at someone who started saving late and feels behind."
Result: specific, opinionated, actionable.

How to write good role prompts

• Be specific. "Senior copywriter who has worked at Stripe and Notion" beats "good writer."
• Add credentials selectively. Years of experience, schools, companies, frameworks they use.
• Add the audience. "Explaining to..." or "writing for..." — both shape the output.
• Combine with a task. Role alone isn't enough; pair it with a clear task and format.

Useful role templates

• "You are a [role] with [N] years of experience."
• "Act as a [role] who has worked at [company type]."
• "Pretend you're a [role] who explains [topic] to non-experts."
• "You are a tough but fair [role] reviewing my [output]."

Related terms

• System Prompt
• Prompt Engineering
• Few-Shot Prompting