Roles & Audience: Who the AI Is and Who It Writes For
Learn how “You are…” role prompts and a clearly described audience change an AI’s tone, vocabulary and level of detail — and where roles stop helping.
On this page 7 sections
Introduction
Ask “What is recursion?” and you will get a reasonable textbook answer. Ask a children’s author to explain it to a ten-year-old, or a senior engineer to explain it to a colleague who is debugging a stack overflow, and you get three very different explanations of the same idea. Roles and audiences let you choose which one you get.
This lesson shows how to use both, and explains why the audience is often the more powerful of the two.
- What a role prompt changes — and what it does not.
- How to describe an audience so the answer fits the reader.
- How to combine role and audience in one prompt.
What a Role Does
A role prompt tells the model which perspective to answer from: “You are an experienced high-school chemistry teacher”, “You are a careful code reviewer”, “You are a friendly customer-support agent for an online bookshop”. In apps built on AI, this often goes in the system prompt — a set of instructions the app sends before your message — but you can also simply start your own message with it.
A role mainly changes three things: vocabulary (technical or plain), priorities (a reviewer looks for problems, a coach looks for strengths) and tone (formal, encouraging, blunt). It is a quick way to set many small expectations at once.
Roles are especially useful for feedback. “Review my essay” tends to produce polite praise. “You are a strict university examiner. Point out the three weakest arguments and explain why” produces criticism you can act on.
Writing for an Audience
The reader matters as much as the writer. Describing the audience tells the model which words it can assume are known, how much background to include, and what the reader cares about. Be concrete: “a hospital administrator with no technical background who must approve the budget” is far more useful than “a non-technical person”.
Good audience descriptions often mention what the reader already knows, what they want to do with the answer, and how much time they have. A one-line summary for an executive and a step-by-step guide for a new employee can describe the same project, but they are completely different documents.
A prompt says: “Explain what an API is to a small-restaurant owner who wants to accept online orders. 3 sentences.” Which reply best matches the audience?
Combining Role and Audience
The strongest prompts usually set both: who is speaking and who is listening. Together they define expertise, tone and level of detail in a sentence or two.
You are a patient computer science tutor.Explain recursion to a 15-year-old who knows basic loopsin Python but has never seen a function call itself.Use one short code example and a real-life comparison.Keep it under 200 words.Notice how the audience line does the heavy lifting: it says what the reader already knows (loops), what is new to them (functions calling themselves) and how to bridge the gap (a code example plus a comparison). The role adds patience and a teaching mindset.
Which rewrite makes this prompt most likely to produce a useful explanation for a parent helping with homework?
You are a genius mathematician. Explain fractions.
Where Roles Stop Helping
- A role does not add knowledge. “You are a doctor” does not give the model medical facts it lacks, and it does not make answers more reliable — check important facts with real experts and sources.
- Exaggerated roles (“world’s best…”, “genius”) rarely help; describing the task and reader precisely helps far more.
- A role cannot override missing context. A “senior lawyer” still needs to know which country’s law applies.
- Conflicting signals confuse the model — for example a formal legal role combined with a request for a jokey social-media tone.
Key Takeaways
- A role sets vocabulary, priorities and tone in one sentence.
- The audience description is often more powerful: what the reader knows, wants and needs.
- Combine both for explanations that fit the reader.
- Roles change style, not knowledge — verify facts that matter.
Summary
Roles tell the model how to speak; the audience tells it who it is speaking to. Next you will learn an even more direct way to show the model what you want: giving it examples.