Here’s a quick experiment.
Ask AI: “How do I negotiate a raise?”
You’ll get decent advice. Generic, but fine.
Now try: “You’re a career coach who has helped 500+ professionals negotiate salary increases. How should I negotiate a raise?”
Different response, right? More specific. More confident. More like talking to someone who actually knows what they’re doing.
This is role prompting. And it’s probably the simplest way to dramatically improve AI output.
What Role Prompting Does
When you give AI a role, you’re not just being creative. You’re activating a different pattern of responses.
AI models are trained on text from millions of sources—including content written by experts, professionals, teachers, and specialists of all kinds. When you say “You’re a senior developer,” the model draws from patterns associated with how senior developers communicate, what they know, and how they approach problems.
It’s like tuning a radio. The knowledge is all there. The role tells AI which frequency to broadcast on.
The Basic Format
Role prompting is simple:
You are a [role]. [Your request].
That’s it. You can elaborate, but even this basic format works.
Examples:
You are a senior Python developer. Review this code for potential bugs.
You are a patient math tutor who specializes in helping adults
overcome math anxiety. Explain how percentages work.
You are a direct-response copywriter with 20 years of experience.
Write a headline for this product.
Why It Works So Well
Role prompting succeeds because it solves several problems at once:
1. It Sets Tone and Style
A “friendly customer service rep” writes differently than a “technical documentation specialist.” The role implicitly communicates how formal, detailed, or warm the response should be.
2. It Defines Expertise Level
“You’re a beginner-friendly cooking instructor” produces different explanations than “You’re a professional chef.” The role tells AI how much to assume about the reader’s knowledge.
3. It Focuses the Response
When AI is a “financial advisor,” it approaches questions through a financial lens. When it’s a “therapist,” the same question gets a different treatment. The role creates a consistent perspective.
4. It Unlocks Specialized Knowledge
AI has information about many fields. A role helps surface the relevant expertise. “You’re a contract lawyer” brings legal patterns forward; “You’re a marketing strategist” brings marketing patterns.
10 High-Impact Roles to Try
Here are roles I use regularly, with examples:
1. The Expert Reviewer
You're a senior [field] professional reviewing work from a junior
colleague. Be constructive but thorough. Point out issues I might
have missed.
2. The Teacher
You're a patient teacher who's great at explaining complex topics
to beginners. Use analogies and simple language. Check understanding
as you go.
3. The Devil’s Advocate
You're a critical thinker whose job is to find flaws in ideas.
Challenge my plan. What could go wrong? What am I not considering?
4. The Experienced Practitioner
You're someone who has done this exact task hundreds of times.
What mistakes do beginners make? What shortcuts actually work?
What should I watch out for?
5. The Target User
You're a [description of target audience]. React to this [product/
content/idea] honestly. What appeals to you? What doesn't? What
questions do you have?
6. The Editor
You're a professional editor who values clarity and concision.
Your job is to make writing better, not different. Cut the fluff
but preserve my voice.
7. The Strategist
You're a strategic advisor who thinks three moves ahead. Don't just
answer my immediate question—help me see the bigger picture and
potential consequences.
8. The Explainer
You're a science communicator who makes complex topics accessible.
Explain [topic] so a smart 12-year-old would understand. No jargon
unless you explain it.
9. The Interviewer
You're a skilled interviewer preparing me for a [type] interview.
Ask me challenging questions one at a time. Give feedback on my
answers. Be tough but fair.
10. The Systems Thinker
You're a systems analyst who looks for patterns and connections.
Help me understand how the pieces fit together and where the
leverage points are.
Making Roles More Specific
The more specific your role, the better your results.
Generic:
You're a writer. Write a product description.
Better:
You're a copywriter who specializes in SaaS products. You write
descriptions that are clear, benefit-focused, and avoid buzzwords.
Write a product description for [product].
Even better:
You're a B2B SaaS copywriter who's studied Basecamp's writing style.
You believe in clarity over cleverness, benefits over features, and
short sentences over long ones. Write a product description for
[product] aimed at small business owners who are frustrated with
complicated software.
Add details that matter:
- Experience level: junior vs senior, 5 years vs 20 years
- Specialization: “marketing copywriter” vs “direct-response copywriter”
- Style: “who favors concise communication” vs “who writes in an engaging, narrative style”
- Values: “who prioritizes user privacy” or “who believes in radical honesty”
Combining Roles With Other Techniques
Role prompting works even better when combined with other methods:
Role + Task + Format
You're a UX researcher [role]. Analyze this user interview transcript
[task]. Present findings as bullet points grouped by theme, with direct
quotes as evidence [format].
Role + Constraints
You're a senior developer [role]. Review this code [task]. Focus only
on security issues [constraint]. Explain each issue so a junior
developer could understand [constraint].
Role + Few-Shot Examples
You're a customer service rep who handles complaints with empathy
and solutions [role].
Example:
Customer: "This is the third time my order was wrong!"
Response: "I completely understand your frustration—three mistakes
is unacceptable. Let me fix this right now and make sure it doesn't
happen again. [solution details]"
Now respond to:
Customer: "I've been waiting 45 minutes for support!"
Response:
When Roles Don’t Help
Role prompting isn’t magic. It’s less effective for:
- Pure factual questions: “What year was the Eiffel Tower built?” doesn’t benefit from a role.
- Tasks where you want neutrality: If you need balanced analysis, a strong role might bias the response.
- Very specific technical tasks: Sometimes clear instructions beat persona framing.
Also, research shows mixed results for certain role types. Very generic roles (“act as an assistant”) add little. Stereotyped roles can produce biased output. The most effective roles are specific and task-relevant.
The Secret: Roles Are Instructions in Disguise
Here’s what’s really happening.
When you say “You’re a patient teacher who uses simple language,” you’re actually saying “Be patient, use simple language.” The role format just makes it more natural and comprehensive.
A good role bundles many implicit instructions:
- Tone (professional, friendly, casual)
- Expertise level (what to assume)
- Communication style (concise, detailed, narrative)
- Priorities (what to focus on)
You could spell all this out explicitly. But “You’re a senior developer” is faster and often more effective than listing twelve different instructions.
Your Role Prompting Template
Here’s a template to start with:
You are a [specific role] with [relevant experience/expertise].
Your communication style is [tone description].
You prioritize [key values or focus areas].
[Your actual request/question]
Customize based on what matters for your task. Not every prompt needs all these elements—but knowing what you can specify helps you tune the output.
Try It Now
Take something you asked AI recently that gave a generic response.
Add a role. Make it specific. See what changes.
I think you’ll find this becomes one of your most-used techniques. It’s simple, it’s fast, and it works for almost everything.
Give AI a role, and watch it rise to the occasion.