How to write better ChatGPT prompts for everyday tasks

What you need

• An AI chat assistant that you can open and type into.
• One task you want help with, such as planning a short trip or writing an email.
• A few notes about the details: who the answer is for, what you need, and how formal or friendly it should sound.

Step by step

1. Decide the one thing you want. Write the task in a single sentence. This helps you notice when a reply has missed the point.

2. Add the context. Include the details the assistant cannot guess, such as the place, dates, budget, or who will read the message. Clear, specific prompts with enough context help the assistant give more accurate and relevant answers.

3. Name the tone. Use a descriptive word such as friendly, formal, professional, humorous, or serious. For example, you could end your request with: ‘Explain this in a friendly and engaging tone.’

4. Send your prompt. Type your request in full sentences and send it. There is no need for special code or commands.

5. Read the reply slowly. Ask yourself whether it does the task you wrote down in step 1. Note what is missing, too long, or off target.

6. Refine instead of starting over. Change the wording, add the context you left out, or simplify the request. Prompting is usually a repeated cycle of asking, reviewing, and adjusting.

7. Steer with one firm sentence. If the reply drifts away from what you wanted, state the behaviour you want in one clear sentence. Newer models follow written instructions closely and literally, so a single clear sentence is often enough to get them back on track.

Example 1: planning a trip (example only)

I am planning a three-day trip to a coastal town in spring on a set budget for food and transport. I enjoy walking and quiet places. Suggest a simple day-by-day plan and explain it in a friendly, relaxed tone.

Example 2: writing an email (example only)

Write a short email to my neighbour asking if I can borrow a ladder on Saturday morning. Keep it polite and friendly, and under 100 words.

If the first draft of the email is too long, you could refine it with: ‘Make this shorter and warmer, and keep the request in the first two lines.’

Tips and common mistakes

• Avoid vague requests. Words like ‘something nice’ or ‘a good plan’ leave the assistant guessing. Be as precise as you can.
• Do not stop at the first reply. Your first prompt is a starting point. Use the reply to see what to add.
• Do not pile everything into one message. A short, clear prompt followed by refinements is often easier to fix than one very long prompt.
• Expect variation. Language models are not fully predictable, so asking the same question again may give a different answer. Keep the version you like and refine from there.
• Do not assume a polite tone means accuracy. A friendly reply can still contain mistakes.

Use it responsibly

• Check every answer. Compare facts, dates, prices, and any health, legal, or money advice with a trusted source before you rely on them.
• Never paste private data. Do not enter passwords, bank details, ID numbers, or personal information about other people into a prompt.
• Follow the rules. Check your workplace or school policies and the service’s own terms before using an AI assistant for work or study.
• Keep your account secure. Use a strong, unique password, and sign out when you use a shared or public device.