How Socrates prompts ChatGPT.

Self-Ask Prompting - or the Socratic way of prompting ChatGPT.

Jun 01, 2025

Steve Jobs predicted ChatGPT in 1985.

“My hope is that someday, when the next Aristotle is alive, we can capture the underlying world view of that Aristotle in a computer. And someday, some student will be able to not only read the words Aristotle wrote, but ask Aristotle a question and get an answer.”

A visionary, ahead of his time.

Especially when I end up falling into a rabbit hole recently of prompt engineering academic papers.

The latest I read (& want us to master) is “Self-Ask prompting, 2022 Press et al.”.

Self-Ask Prompting - or the Socratic way of prompting ChatGPT - is a technique to tells the model to decide whether it needs follow-up questions, ask them aloud, answer them, and only then state the final answer.

This technique slashes “long-but-wrong” hallucinations (you know, those moments when ChatGPT is stuck in loops) from 40% to 17%.

Let’s master it today, step-by-step, without any code or technical terms.

The ‘How’ of Self-Ask Prompting

Use this snippet for your next prompt on ChatGPT:

You must decompose the task before answering.

Question:
<YOUR ACTUAL INPUT = YOUR PROMPT HERE>

Step 1 – Need follow-up questions? Answer Yes/No.

If Yes, loop:
Follow-up #:
<leave blank – you (ChatGPT) write the clarifying question>

Answer #:

(Repeat until no further follow-ups are needed.)

Step 2 – Final output:
Use only the facts in Answer lines. 
If key info is missing, say “Insufficient information.”
 
End with a 0-100 % confidence score.

The only part you have to manually edit is “YOUR ACTUAL INPUT = YOUR PROMPT HERE”. The rest is automatic.

Let’s get you some examples for you to visualize it.

Example 1 – NDA clause for a SaaS reseller deal

You must decompose the task before answering.

Question:
<Draft a two-paragraph non-disclosure clause for a SaaS reseller agreement that covers customer data, survival period, and governing law.>

Step 1 – Need follow-up questions? Answer Yes/No.

If Yes, loop:
Follow-up #:
<leave blank – you (ChatGPT) write the clarifying question>

Answer #:

(Repeat until no further follow-ups are needed.)

Step 2 – Final output:
Use only the facts in Answer lines. 
If key info is missing, say “Insufficient information.”
 
End with a 0-100 % confidence score.

Here’s ChatGPT’s answer on this one:

Why does it work:

You can easily challenge its own thinking.
You can easily add new follow-up questions.
It’s you + ChatGPT making decisions, together.

Example 2 – Blog-post outline for SMB audience

You must decompose the task before answering.

Question:
<Outline a 1,200-word blog post that explains, to non-technical small-business owners, how AI chatbots cut customer-service costs.>

Step 1 – Need follow-up questions? Answer Yes/No.

If Yes, loop:
Follow-up #:
<leave blank – you (ChatGPT) write the clarifying question>

Answer #:

(Repeat until no further follow-ups are needed.)

Step 2 – Final output:
Use only the facts in Answer lines. 
If key info is missing, say “Insufficient information.”
 
End with a 0-100 % confidence score.

Here, the result is actually even better than example #1:

ChatGPT realizes it can’t write a blog without the proper answers.

So it forces you to be clearer, so ChatGPT can write a better blog.

AI + human at its best.

Example 3 – LinkedIn hooks for a new tool

You must decompose the task before answering.

Question:
<Write five LinkedIn hook sentences to announce my new platform "PromptWatch" that helps businesses get mentioned inside ChatGPT, Gemini, Perplexity...>

Step 1 – Need follow-up questions? Answer Yes/No.

If Yes, loop:
Follow-up #:
<leave blank – you (ChatGPT) write the clarifying question>

Answer #:

(Repeat until no further follow-ups are needed.)

Step 2 – Final output:
Use only the facts in Answer lines. 
If key info is missing, say “Insufficient information.”
 
End with a 0-100 % confidence score.

For this, I will show you my step-by-step approach from start to finish.