I’ve written before about how to ask ChatGPT to improve a so-so prompt. But what about those times with ChatGPT, Claude, or Gemini absolutely crushes it, giving you exactly what you needed from your prompt?
Every so often, you’ll strike upon the perfect prompt, one that delivers a surprisingly thorough, incisive, or otherwise effective answer. The trick is learning how to repeat that success.
The first step: Stop. Don’t just move on to a follow-up prompt. Instead, take a beat and ask ChatGPT (or whichever AI chatbot you happen to be using) why your prompt worked so well.
The next time you get a great answer from an AI, try this prompt immediately afterward:
That was exactly what I needed. Two things. First, explain what about my prompt made this work so well. Second, turn my request into a reusable, fill-in-the-blank template I can use next time and tell me which parts matter most.
One thing that sets this prompt apart from others I’ve covered is its simplicity. There’s no coding tags, roleplaying, or “be a panel of experts” stuff here. Instead, it’s a straightforward “get the recipe” prompt.
Also, this prompt is asking for two things that actually work hand-in-hand: an explanation for why the prompt worked and a template for doing it again.
That “create a template” is doing double-duty here, giving you both a blueprint for repeating the prompt’s success and ensuring the AI gives you concrete details, rather than merely flattering you on your prompting prowess.
Here’s an example of a ChatGPT prompt (using GPT-5.5 Instant) that yielded a spot-on response:
Help me write an email asking my landlord to fix the dishwasher. We have a friendly, but slightly tense relationship. He’s slow to respond and I don’t want to come off as demanding, but this is the third time I’ve asked. Keep it short and polite, but with a clear deadline.
When I followed up with the “explain what went right” prompt, here’s a snippet of the answer I got:
What made your prompt work so well is that it specified the decision-making constraints, not just the task.
Most writing requests look like this:
“Write an email asking my landlord to fix the dishwasher.”
That’s enough to produce something grammatically correct, but it leaves a lot of important choices up to the AI.
Your prompt filled in the missing context:
Audience: “my landlord”
Relationship: “friendly but slightly tense”
Behavioral context: “He’s slow to respond.”
Your goal: “I don’t want to come off as demanding.”
Relevant history: “this is the third time I’ve asked.”
Output constraints: “Keep it short and polite but with a clear deadline.”Each of those eliminated an ambiguity.
Then ChatGPT generated the reusable template:
Write a [type of message] to [recipient].
Relationship: [How we normally interact.]
Context: [What happened and any relevant history.]
Goal: [What I want them to do.]
Things to balance: [Examples: friendly but firm, apologetic but confident, professional but warm.]
Constraints: [Length, deadline, tone, anything that must or must not be included.]
Combined, the “why did it work” explanation and “make me a template” request reveal the winning strategy behind the prompt: it focuses on the “what” rather than the “how.”
Try this “get the recipe” prompt with other prompts that worked well for you, and hopefully you’ll discover new pearls of AI prompting wisdom.