2
Top highlight
1
1
Member-only story
Forget every prompt gimmick you’ve ever learned; this is all you need.
Follow
5 min read
Jul 13, 2025
4.8K
114
Listen
Share
More
Zoom image will be displayed
Generated by the author with AI.
I set out to write the ultimate ChatGPT prompt, and I think I've done it.
My favorite article I’ve ever written (The Most Important ChatGPT Prompt) made waves in the AI community 18 months ago.
Ever since the release of that article, I have seen countless posts on social media surrounding the core idea behind the prompt in different ways. The concept is simple: tell ChatGPT to ask you questions before it starts to reply:
Please ask me any questions you have about this before you begin.
This strategy forces ChatGPT to set aside its innate desire to instantly please you and forces you to provide it with more valuable information to complete your request.
The problem is, a lot of these prompting frameworks that I’ve come across are clouded with too much information. Check out this bloated, gross prompt known as “Lyra”:
This is way too much for the AI to handle.
If you read my most recent article, you’ll know that clean, direct information is the only thing that really matters when it comes to prompting these new intelligent models.
That’s why I’ve decided to create the final boss of prompts.
The Ultimate ChatGPT Prompt?
While writing this new prompt, I worked under these constraints to ensure I wrote something maximally useful:
- It has to be less than 25 words or less
- It has to be easily memorizable (so users don’t have to copy and paste it every time)
- It has to be vastly more effective than the standard core of the prompt written above
- It must be optimized for reasoning models (these are the new meta)
- It had to achieve my goal in one try every time
After hours of testing and cross-checking with old approaches, I’ve arrived at the definitive prompt:
Output an overview of every single dimension of my request. Find points of uncertainty. Then, ask me as many clarifying questions as possible.
That’s it.
Guess what? In my testing, I’ve been happy with ChatGPT’s first response nearly every time. Let’s see it in action…
Zoom image will be displayed
Before we get back to the article, please consider trying my new infinite memory voice companion, Amber AI. Your first 30 minutes are on me!
Zoom image will be displayed
After three years of building AI products for my clients, I’ve finally decided to formally launch my own product.
Amber AI has a simple phone call interface, no apps, no BS. You can choose the voice of your liking, and its personality grows with you. Thanks to our proprietary memory system, Amber never forgets anything.
It would mean a great deal to me if you decided to sign up! Please follow up with any feedback you might have!
Back to the story.
Zoom image will be displayed
The Ultimate Prompt — Results
So I’ve dropped the prompt, but the proof is in the pudding.
Let’s check out a side-by-side comparison to a classical workflow with ChatGPT versus using this prompt.
Example: A Video Script
Let’s say that I want to create a video ad for my new product that I just mentioned above (this is hypothetical as I haven’t started advertising yet).
The classical flow is blindly pasting in my request to ChatGPT, and it regurgitates back a bunch of total garbage:
Zoom image will be displayed
It instantly opens up the canvas without any regard for specifics behind my request.
As you can see, this isn’t very helpful. It just assumed a bunch of stuff that I didn’t tell it, and it only has a superficial understanding of my product and how I intend to sell it.
Enter the ultimate prompt:
Zoom image will be displayed
A much better start, it doesn’t assume anything yet…
A very deep analysis follows this, and then questioning from the model:
A very large analysis of the prompt, followed by a bunch of questions to help it along.
As you can see, it begins by analyzing the request from all angles, a process known as context rooting or step-back prompting. This helps models “think through” requests more comprehensively, while being cordoned to the richer semantic space of your request.
What happens after you answer all the questions?
Well, in this case, it asked me 33 questions (hey, I never said this prompt didn’t involve any work…) and after answering them, it gave me a deeply relevant and insightful answer to my request…
Zoom image will be displayed
And here we can see the much better, structured to my liking output of this request.
Now the output is not only structured how I wanted it, but rooted exactly as I envisioned it.
After using this prompt a bunch over the past few days, I’ve it to be much more performant than the original “ask me any questions you may have” prompt.
I also see improved performance over larger-scale prompting frameworks that bog down the AI’s context.
The Benefits of the Ultimate Prompt
Here is a short list of the benefits of using this prompt:
- Increases your understanding of your request (chances are, it will think of things that even you didn’t consider)
- Forces you to give it all the background information it needs
- Roots its context better, making it understand the breadth of possible outputs
- Adds more length to the chat, increasing the likelihood of a more comprehensive, “high effort” response from the AI.
I hope this is helpful for your workflow! Please tell me about your use cases for this prompt in the comments; I’m excited to see how y’all apply it.
Thanks for reading!
-Jordan
Made with Bullet