Many great prompts start with a role: “Act as an experienced recruiter” or “You are a patient maths tutor.” It can look like a gimmick, but a role does real work. It tells the AI which knowledge to draw on, what standards to apply and how to talk to you.
What a role changes
- Expertise: a “senior tax accountant” will focus on different details than a general assistant.
- Tone: a “friendly kindergarten teacher” explains very differently from a “university professor”.
- Standards: a “meticulous copy editor” is more likely to catch small errors.
Make the role specific
“Act as a marketer” is vague. Compare:
Act as a B2B SaaS marketer with ten years of experience writing landing pages for technical buyers who are sceptical of hype.
The second version implies an audience, a style and a set of priorities, without you spelling each one out.
Add the audience
A role works best when paired with who it's serving: “Explain this as a doctor would to a nervous patient with no medical background.” The role controls knowledge; the audience controls language.
Combine roles for better reviews
Ask the AI to review its own work from a second perspective: “Now review this landing page as a sceptical customer. What would make you leave?” Switching roles is a quick way to find weaknesses.
Know the limits
A role doesn't give the AI knowledge it doesn't have, and it doesn't make answers more reliable on its own. For legal, medical or financial decisions, use AI to understand the topic and prepare questions, then confirm with a qualified professional.