If you’re familiar with basic prompt engineering techniques and want to learn advanced prompt engineering techniques, or if you find the basic methods somewhat insufficient for more complex problems, this article is for you.
In this two-part series, we’re introducing two advanced techniques—Self-Ask Prompting and Generated Knowledge Prompting—that you can add to your toolkit to further optimize your interactions with ChatGPT and tackle more complex problems.
This article covers the first technique: Self-Ask Prompting. Be on the lookout for the next one, where we will discuss Generated Knowledge Prompting.
Table Of Content
What is Self-Ask Prompting
AI models like ChatGPT could sometimes struggle with complex, multi-faceted, or multi-step reasoning, and as a result, return incomplete, vague, or incorrect responses. They tend to answer questions directly without breaking them down into smaller, manageable parts. This will lead to oversimplified outputs.
If you are familiar with basic prompt engineering techniques, the Chain of Thoughts (CoT) technique might come to your mind as a solution for such an issue. However, you might also consider that CoT is particularly suited for tasks that involve logical or sequential reasoning—especially when each step builds on the previous one. For example, you can apply it with a prompt like:
What is the best way to travel from New York City to Brooklyn? Let's go step-by-step.
While CoT works well for such step-by-step tasks, it might not be the best method for complex tasks that are multi-faceted in nature—those that require deep exploration, critical analysis, or understanding of the problem from multiple aspects.

For example, when you ask ChatGPT, “How can companies improve employee productivity?” and want it to dissect this from various angles (rather than listing sequential steps), you can apply the Self-Ask Prompting technique.
How Does Self-Ask Prompting Work?
When you assign a task to the AI and apply Self-Ask Prompting, the AI breaks the main task into manageable pieces by generating “sub-questions”, i.e., smaller and more specific questions related to the main task. It then answers these sub-questions, guiding itself through the process of understanding and solving the problem.
This approach mirrors how humans often try to figure things out: We look at a problem, self-ask a series of related questions, and then self-answer them in order to understand the issue and find a solution—which is exactly the process we want the AI to model.
We/the AI would consider the problem from multiple perspectives during the process. This is why this approach is particularly effective for problems with different dimensions.
With the right prompt, the AI’s self-asking process is self-guided, meaning you can sit back and observe as it questions and answers itself. This advanced prompt engineering technique drives the AI to dissect the issue step-by-step, and will often lead to a clearer and more precise final answer.
Steps to Apply Self-Ask Prompting
- Define the Main Task: In your prompt, clearly state the problem or task you want the AI to solve.
- Craft a Self-Ask Prompt: Include instructions for the AI to generate relevant sub-questions to explore different aspects of the main task.
- Set Clear Expectations: Specify how detailed the sub-questions and answers should be to keep the AI focused and avoid irrelevant digressions.
- Run the Prompt: Execute the prompt and allow the AI to self-generate and answer sub-questions step-by-step.
- Evaluate and Refine: Review the AI’s output, checking for relevance and accuracy. Adjust the prompt if necessary to guide the AI towards a better understanding of the task.

Examples
To demonstrate how to apply Self-Ask Prompting, let’s look at an example using the question: “What should I consider when introducing a new cat to my home that already has pets?”
This is the prompt you might send to ChatGPT:
You are an AI that helps solve complex problems by breaking them into smaller, manageable sub-questions. For the task of introducing a new cat to a home that already has pets, follow these steps:
1. Generate relevant sub-questions that explore different aspects of this situation.
2. Answer each sub-question in detail to provide a comprehensive understanding of what to consider.
3. Use the answers to form a clear, actionable plan.
Start by asking the first sub-question.
Noticed how we explicitly instruct ChatGPT to:
- Break the main problem into smaller, manageable sub-questions
- Explore different aspects
- Answer the sub-questions
These are the key elements that define Self-Ask Prompting.
The result was a series of 6 self-asked and self-answered questions that covered different aspects (snapshot of the output below)…

…and ChatGPT’s recommendation based on the aspects it considered (snapshot below):

Conclusion
Self-Ask Prompting is a powerful and advanced prompt engineering technique that you’ll want to add to your tool bag for handling complex, multi-faceted problems.
It prompts the AI to reason through the issue from multiple angles rather than providing a surface-level answer. This way, the AI gains a more well-rounded understanding of the task and can provide you with more precise and comprehensive answers. Plus, you might learn some insights from observing the AI’s self-questioning and self-answering process!
Coming up next is the second technique we’ll introduce to you: Generated Knowledge Prompting.
Meanwhile, try out the Self-Ask Prompting technique. If you have any questions or comments, feel free to leave them in the comments section. Happy prompting!
No Comment! Be the first one.