I’ve previously discussed how to ask ChatGPT to refine a mediocre prompt. But what about those moments when ChatGPT, Claude, or Gemini nail it, providing exactly the response you were aiming for?
Occasionally, you’ll stumble upon a prompt that yields an exceptionally detailed, insightful, or effective answer. The key is learning how to replicate that success consistently.
The first step: pause. Instead of immediately crafting a follow-up prompt, take a moment to ask your AI (whether ChatGPT or another) why that particular prompt worked so well.
Next time you receive an outstanding reply from an AI, follow up with this request right after:
“That was exactly what I needed. First, explain what about my prompt made it so effective. Second, turn my request into a reusable, fill-in-the-blank template I can use later, and tell me which parts are most important.”
This prompt stands out because of its clarity and simplicity—there’s no need for complex coding tags, role-playing instructions, or expert impersonations. It’s a straightforward way to extract “the recipe” behind a good prompt.
Importantly, this request asks for two linked outputs: an explanation of why the prompt succeeded and a reusable template for future use. The “create a template” component serves a dual purpose—it provides you with a blueprint for repetition and prompts the AI to give specific, actionable details, rather than just complimenting your prompting skills.
Here’s an example I used with GPT-5.5 Instant that produced an excellent answer:
“Help me craft an email asking my landlord to repair the dishwasher. We have a friendly but slightly tense relationship. He’s slow to respond, and I want to avoid sounding demanding, but this is the third time I’ve asked. Keep it short and polite with a clear deadline.”
When I asked the AI to explain what worked, it responded:
“Your prompt was effective because it outlined the decision-making constraints, not just the task.”
Most requests like “Write an email to my landlord about fixing the dishwasher” only specify the task. While that can generate a grammatically correct output, it leaves many key choices up to the AI.
Your prompt added crucial context:
- Audience: “my landlord”
- Relationship: “friendly but slightly tense”
- Behavior: “He’s slow to respond”
- Goal: “I don’t want to come off as demanding”
- History: “this is the third time I’ve asked”
- Format constraints: “Keep it short and polite, with a clear deadline”
Each element clarifies the ambiguity, guiding the AI toward a more precise response.
The AI then produced a reusable template:
Write a [type of message] to [recipient].
Relationship: [How we normally interact.]
Context: [What happened and relevant history.]
Goal: [What I want them to do.]
Tone/Balance: [Examples: friendly but firm, apologetic but confident.]
Constraints: [Length, deadline, tone, inclusions/exclusions.]
Together, the explanation of why it worked and this template form reveal a core strategy: focusing on the “what” rather than the “how.” It emphasizes specifying the context and constraints, which leads to better results.
Try this “get the recipe” approach with other successful prompts, and you may uncover additional insights into crafting effective AI instructions.