How to Write Better AI Prompts: 8 Techniques That Work

A "prompt" is simply the message you send to an AI chatbot. The quality of the answer depends heavily on the quality of your request. If you want the basics and more examples first, read What is a prompt? The good news: you do not need secret formulas. A handful of habits will noticeably improve results with any major chatbot. Here are eight techniques, each with an example.

1. Say who the answer is for

The same topic needs different explanations for a ten-year-old, a busy manager, and a specialist. Tell the chatbot the audience.

Weak: "Explain inflation."
Better: "Explain inflation to a high-school student in about 150 words, using an everyday example about groceries."

2. State the task and the outcome you want

Be explicit about what to produce: a list, an email, a table, a plan, a comparison. Vague requests lead to vague answers.

Weak: "Help me with my trip."
Better: "Create a three-day itinerary for a first visit to Chicago for two adults who like food and architecture. Present it as a day-by-day list with morning, afternoon and evening."

3. Give context and constraints

The chatbot only knows what you tell it. Include relevant background, such as your goal, your experience level, your budget limits (described in your own terms), the length, tone or format. Constraints make answers more usable.

• Length: "under 200 words," "five bullet points"
• Tone: "friendly but professional," "plain and direct"
• Exclusions: "avoid jargon," "do not suggest anything that needs a car"

4. Assign a role (when it helps)

Asking the chatbot to act as a particular kind of helper can set the style and focus. For example: "Act as a patient writing tutor. Review my paragraph and suggest three specific improvements, explaining why." A role is not magic, and it does not give the model real credentials, but it does steer tone and structure.

5. Show an example of what you want

Examples are one of the most reliable ways to get the right format. Paste a short sample and say "match this style" or "use this format."

Example: "I write product descriptions like this: [paste your sample]. Write a description for a stainless steel water bottle in the same voice."

6. Break big tasks into steps

Instead of asking for a complete report in one go, work through it: first ask for an outline, then refine the outline, then ask for each section. You stay in control and catch problems early. You can also ask the chatbot to "think through this step by step" or to list assumptions before giving the final answer.

7. Iterate: treat the first answer as a draft

Very few people get a perfect result on the first try. Follow up with specific feedback:

• "Make it shorter and remove the second paragraph."
• "This is too formal. Rewrite it as if talking to a friend."
• "Give me three alternatives with different openings."
• "What did you assume that I did not tell you?"

Because the chatbot remembers the earlier messages in the same conversation, refining is usually faster than starting from scratch.

8. Ask it to check itself, then verify yourself

You can ask: "List any claims in your answer that you are unsure about," or "Which parts should I verify with a reliable source?" This does not guarantee accuracy, because models can be confidently wrong, but it often surfaces weak spots. For anything important, confirm facts independently. Our guide to checking AI answers for mistakes explains how.

A reusable prompt template

You can adapt this structure for almost any request:

Goal: [what you want to accomplish]
Audience: [who it is for]
Context: [relevant background]
Task: [exactly what to produce]
Format: [list, table, email, word count]
Tone: [e.g., warm, formal, concise]
Avoid: [things to leave out]

Before-and-after: a full example

Before: "Write an email about the meeting."

After: "Write a short email (under 120 words) to my team reminding them that Thursday's project meeting has moved from 10:00 to 11:30. Keep the tone friendly and ask them to reply only if the new time is a problem. End with my name, Sam."

The second prompt supplies the audience, facts, length, tone and the desired action. The chatbot has far less guesswork to do.

Common mistakes to avoid

• Being too vague. "Make it better" gives the model nothing to aim at. Say what "better" means.
• Cramming unrelated requests together. Separate tasks usually get better results.
• Pasting sensitive data. Do not include passwords, financial account numbers or private details. Read our privacy guide.
• Trusting every detail. Treat names, dates, quotes and statistics as claims to verify.

Practice ideas

1. Take an email you wrote and ask the chatbot to rewrite it three ways: shorter, warmer, and more formal.
2. Ask it to turn rough notes into a checklist.
3. Ask for an explanation of something you are learning, then ask it to quiz you. See using AI to study.

New to all of this? Start with what an AI chatbot is.

See these techniques applied to real tasks: job-search prompts in AI tools for job seekers and listing-description prompts with guardrails in AI for real estate listing descriptions.

Key takeaways

• Specific, contextual prompts beat vague ones.
• Tell the AI the audience, task, format and tone.
• Use examples and break big jobs into steps.
• Iterate, and verify anything that matters.

Prompt examples for educators are in AI for lesson planning (part of the AI for Teachers hub).

Prompting for your job? See the task guides in AI for Nurses, AI for Small Business Owners, AI for Plumbers and Trades and AI for Construction Company Owners, each with copy-paste prompts.