How to write effective prompts for AI assistants

What you need

To follow this guide, you only need access to an AI assistant, such as ChatGPT or a model from the GPT-4.1 family.

Step by step

1. Create a clear and specific initial prompt Start by writing an instruction that is precise and avoids being vague. Providing enough context helps the model understand exactly what you are looking for.

*Example: Instead of saying “Write about travel,” try “Write a three-day itinerary for a family trip to London focusing on museums and parks.”

2. Add descriptive adjectives for tone If you want the response to sound a certain way, use specific words to describe the desired personality or style. This helps guide the model’s output.

*Example: You can use words like formal, informal, friendly, professional, humorous, or serious. For instance, you might use the prompt: Explain how a camera works in a friendly and engaging tone.

3. Review the AI’s response Once the model provides an answer, read through it carefully to see if it met your expectations and followed your instructions.

4. Refine your prompt through iteration If the first answer isn’t quite right, do not give up. Treat prompting as an ongoing process. You can adjust your wording, add more context, or simplify your request to improve the next result.

5. Use a single sentence to correct behavior If the model’s behavior is different from what you expected, you can often steer it back on course by adding a single, firm, and clear sentence that explains exactly how you want it to behave.

6. Induce step-by-step planning For more complex tasks, you can ask the model to produce an explicit, step-by-step plan. This encourages the model to “think out loud” before providing a final answer, which can help with accuracy.

*Example: If you are planning a small project, you might use a prompt like: Create a step-by-step plan to help me organize my digital photo library.

Tips and common mistakes

• Be precise: Avoid ambiguity. The more specific your instructions, the more relevant the response will be [1].
• Use examples: Providing context examples can help the model understand the task better [2].
• Don’t expect perfection immediately: Prompt engineering is an empirical process, meaning it often requires trying different versions to see what works best [2].
• Avoid manual tool injection: If you are a developer using tools, it is better to use the designated tools field rather than manually typing descriptions into a prompt to minimize errors [2].
• Use sections for complexity: If you need to provide complex examples for a tool, consider creating a specific “# Examples” section in your prompt to keep things organized [2].

Use it responsibly

Always remember to check the answers provided by the AI for accuracy, as they may not always be correct. Protect your privacy by never pasting sensitive or private data into your prompts. Finally, ensure you are following the rules of your workplace or school, as well as the service’s own terms of use.