Same question, wildly different answers — the secret is in the prompt ⓒ Unsplash
The formula for writing prompts that get you exactly the answer you need.
Two people type “write me an email” and one gets a perfect draft while the other keeps hitting regenerate in frustration. The difference isn’t luck — it’s the prompt (how you phrase your request). Here’s the formula for writing prompts that get you close to the perfect answer on the first try.
1. The 4 Elements of a Great Prompt
① Role — tell the AI what perspective or expertise to respond from
② Context — provide the background and purpose specifically
③ Request — state exactly what you want
④ Format — specify the shape of the output (length, tone, structure)
Include these four elements and most of your prompts will dramatically improve.
2. Bad Prompt vs. Good Prompt
“Write me an email”
✅ Strong Example
“You’re a marketing team lead. Write an email to a client contact proposing a meeting time for next week. Keep the tone professional but not overly formal, in 3–4 short sentences. Ask if Tuesday or Thursday afternoon works for them.”
3. Prompt Templates by Situation
📊 Writing a report
“You’re a [role] expert. Draft a report on [topic]. The audience is [who], and the purpose is [purpose]. Lead with the conclusion first.”
📧 Writing an email
“Write an email to [recipient] with the goal of [purpose], in a [tone] tone. Keep it around [length].”
📝 Brainstorming ideas
“Suggest [number] ideas for [topic]. Summarize the pros and cons of each in one line.”
4. Refining the Answer With Follow-Up Prompts
ChatGPT isn’t designed to nail a perfect answer in one shot — it’s a tool for iterative refinement through conversation.
💬 “Make this more concise”
💬 “Rewrite this in a more professional tone”
💬 “Add one concrete example”
💬 “Check if there’s a gap in this logic”
Chaining follow-up requests like these gets you noticeably closer to what you actually want with each turn.
5. Common Prompt Mistakes
| Mistake | Result | Fix |
|---|---|---|
| Vague, one-line question | Generic, bland response | Add specific background and purpose |
| Bundling multiple requests together | Some requests get missed | Number the requests clearly and separately |
| Not specifying the desired format | Output shape doesn’t match expectations | State it explicitly: “put this in a table,” “summarize in 3 lines” |
| Starting a fresh chat instead of following up | Loses context, requires re-explaining from scratch | Continue refining in the same conversation thread |
6. The Power of Role-Play Prompts
Assigning a specific role (expert, editor, coach, etc.) shapes the tone and expertise of the response accordingly. “You’re a copywriter with 10 years of experience” produces far more polished marketing copy. “You’re a fifth-grade teacher” produces a much simpler explanation. Same question, dramatically different — and more useful — output.
The key isn’t length — it’s clarity. Instead of stuffing your prompt with unnecessary qualifiers, keep it concise while including the four core elements: role, context, request, and format.
How to read and analyze PDF files with ChatGPT.
Frequently Asked Questions
Q. Do prompts written in English get better results?
Some people report slightly more refined responses with English prompts in certain contexts, but you can absolutely get excellent results in other languages too. If your goal is a document in a specific language, it’s usually more natural to prompt in that language.
Q. It’s tedious to repeat the same setup every time.
If you have a role or tone you use frequently, set it up once using “Custom Instructions” — then you’ll get consistently styled answers without re-explaining your preferences every session.
Q. Where can I find more prompt examples?
The most effective approach is experimenting directly in conversation to see what phrasing gets the best results for your specific needs. We’ll also continue sharing situation-specific prompt templates throughout this series.
답글 남기기