5 Expert Tips for Creating the Perfect AI Prompt

Explore 5 expert tips for creating the perfect AI prompt, including the key concepts, practical examples, important considerations, and useful guidance.

1. Keep the prompt simple and get straight to the point

Although it's often advertised that you can speak with LLMs in natural or everyday language, that's primarily to make the product appealing to the average user. For better results, you should keep the prompt concise and easy to understand, focusing on action verbs. Every superfluous word or complex phrase you add has the potential to lead to misunderstanding.

Instead of telling your problem to the AI as a story, focus directly on the answer you need. Here are some examples to help you understand better:

Prompt is not good :

AI prompt: "Help me with 1. keep the prompt simple and get straight to the point. for 5 Expert Tips for Creating the Perfect AI Prompt. Provide clear steps, examples, and practical recommendations in US English."

Good prompt :

AI prompt: "Help me with 1. keep the prompt simple and get straight to the point. for 5 Expert Tips for Creating the Perfect AI Prompt. Provide clear steps, examples, and practical recommendations in US English."

2. Assign roles to AI first

For specialized requests, it's best to ask the AI to act as an expert in a specific field, such as 'act as a tour guide' or 'you are a kindergarten teacher'. Because large language models (LLMs) learn from vast textual repositories, this will force them to stick to a particular occupation or personality. The model will use the tone and vocabulary characteristic of the role, limiting itself to the scope of that role, which will prevent it from straying off-topic.

Here are some examples of requirements:

Prompt is not good:

AI prompt: "Help me with 2. assign roles to ai first. for 5 Expert Tips for Creating the Perfect AI Prompt. Provide clear steps, examples, and practical recommendations in US English."

Good prompt:

AI prompt: "Help me with 2. assign roles to ai first. for 5 Expert Tips for Creating the Perfect AI Prompt. Provide clear steps, examples, and practical recommendations in US English."

3. Provide examples

Whenever possible, try to provide multiple examples of how you want the output to look. This will ensure the model delivers exactly what you want without deviating from the goal. These concepts are known as one-shot prompts and few-shot prompts. It's best to provide at least 3 to 5 examples if possible.

Of course, examples aren't necessary for most prompts, but whenever you need to follow a specific structure, they significantly improve results. For example, if you're giving it ratings to understand emotions, you can provide exactly how it needs to categorize those ratings, such as negative, neutral, or positive.

4. Try the Step-Back Prompting method

Step-Back Prompting means dividing a prompt into two steps by asking a principle-level question first, then providing a reverse answer to obtain a detailed response. Essentially, the first question activates the model's background knowledge on the topic, helping the model to provide a better answer when you ask the actual question using the details from the first question.

This approach is better than directly asking for the answer, as it 'encourages LLMs to think critically and apply their knowledge in new and creative ways' .

To help you understand better, the article provides an example illustrating this approach:

First prompt :

AI prompt: "Help me with 4. try the step-back prompting method. for 5 Expert Tips for Creating the Perfect AI Prompt. Provide clear steps, examples, and practical recommendations in US English."

After receiving the answer, use the second prompt (the step-by-step hint).

Second prompt :

AI prompt: "Help me with 4. try the step-back prompting method. for 5 Expert Tips for Creating the Perfect AI Prompt. Provide clear steps, examples, and practical recommendations in US English."

5. Focus on guidance rather than restrictions

LLMs work better when you instruct them to 'do something' rather than tell them 'don't do something'. Instead of telling the model what it should avoid, it's better to specify exactly what to include. If you add restrictions to the question, the model might start guessing what's allowed and may restrict itself more than necessary. Using direct instructions leads to more creative answers.

Although the paper states that using constraints when necessary is acceptable, it's better to use positive guidelines. Here's an example of this concept:

Prompt is not good:

AI prompt: "Help me with 5. focus on guidance rather than restrictions. for 5 Expert Tips for Creating the Perfect AI Prompt. Provide clear steps, examples, and practical recommendations in US English."

Good prompt:

AI prompt: "Help me with 5. focus on guidance rather than restrictions. for 5 Expert Tips for Creating the Perfect AI Prompt. Provide clear steps, examples, and practical recommendations in US English."

These are some helpful tips for everyday use. However, there are many other ways to create better AI responses besides just creating better prompts, such as personalizing the AI.