How to Write Effective AI Prompts (With Copyable Templates)
A good prompt tells the AI four things: what role to play, what task to do, what context matters, and what format you want. Most disappointing answers come from skipping one of these — not from the model being "unable to help."
This guide gives you a repeatable formula and templates you can copy and adapt.
The core formula: role + task + context + format
- Role — who the AI should act as.
- Task — the specific outcome you need.
- Context — background, audience, source material, or constraints.
- Format — bullets, table, email, steps, word count.
Compare these two:
- Weak: "Help me with an email."
- Strong: "Act as a customer-success manager. Write a 120-word email telling a customer their order is delayed by three days. Keep the tone apologetic and professional, and offer a 10% discount."
The second prompt leaves almost no room for a generic answer.
Five habits that improve every prompt
- Be specific about length and audience. "Explain to a complete beginner" and "explain to an engineer" produce different answers.
- Name the format. Ask for a table, numbered steps, or bullet points.
- Provide source material. Paste the text you want used and say "answer only from this."
- Set constraints. Deadlines, tone, things to avoid, and required points all sharpen output.
- Iterate. Follow up with "shorter," "more formal," or "remove the jargon" instead of starting over.
Copyable templates
1. Explanation template
Act as a patient tutor. Explain [concept] to a complete beginner.
Use one everyday analogy, then give the technical definition.
End with three key points and a short FAQ. Avoid jargon.
2. Summarization template
Summarize the text below in five bullet points.
Separate facts from opinions, and note anything that looks uncertain.
Do not add information that is not in the text.
Text: [paste here]
3. Email template
Write a [tone] email of about [number] words to [audience].
Goal: [what you want the recipient to do].
Include these points: [list]. Avoid pushy language and absolute claims.
4. Comparison template
Compare [option A] and [option B] in a table.
Use these dimensions: price, strengths, weaknesses, best-for user.
Then recommend one for [specific situation] and explain the trade-off.
Stay neutral and note where evidence is missing.
5. Brainstorming template
Generate 10 ideas for [goal] aimed at [audience].
For each, give a one-line description and rank them by effort required.
Avoid generic suggestions and mark the three strongest.
6. Editing template
Edit the text below for clarity and natural tone.
Keep the meaning and my voice. Explain each change in a short list.
Text: [paste here]
Advanced moves
- Ask for alternatives. "Give me three versions, from casual to formal."
- Ask the model to challenge you. "List the weaknesses in my plan and any assumptions I'm making."
- Use step-by-step reasoning. For complex tasks, ask it to reason through the problem before giving a final answer.
- Add an uncertainty rule. "If you are unsure, say so explicitly and tell me what to verify."
- Chain prompts. Break a large job into research, outline, draft, and review passes instead of one giant request.
Common mistakes
- One-word prompts. They give the model nothing to work with.
- Asking multiple unrelated things at once. Split them into separate questions.
- Forgetting the audience. The same content written for a boss and a customer should look different.
- Accepting the first draft. Prompting is a conversation; refinement is where quality appears.
Frequently asked questions
Is there a single perfect prompt? No. Models and products change, so a prompt that works today may need adjustment later. Reliable structure matters more than a magic phrase.
Should I use long or short prompts? As long as needed to carry the role, task, context, and format — no longer. Padding does not help.
Can I use the same prompt in different languages? You can translate the structure, but write it in the target language and respect local phrasing rather than converting word by word.
Key takeaways
- Use role + task + context + format every time.
- Name the audience, length, and output structure.
- Reuse the templates and refine through follow-up prompts.
- Treat the first answer as a draft to iterate on.
Next, read our beginner guide to ChatGPT and our explainer on AI hallucination and how to verify answers.
Sources
Learn by doing
Read a guide, then practice with bite-sized AI lessons you can take offline.
Start learning free