Use the "Feynman Technique" to Evaluate the Quality of Your AI Prompts

You write a prompt, and the AI gives you an answer—but how do you know if that answer is any good?

Most people rely on intuition: "It looks okay." But intuition is the most unreliable quality control tool. Today, I’m sharing a proven method: reverse-engineer your prompt evaluation using the Feynman Technique.

What is the Feynman Technique?

The physicist Richard Feynman had a habit: whenever he learned a new concept, he would try to explain it in the simplest possible language to someone who knew nothing about the subject. If he couldn’t explain it clearly, it meant he didn’t truly understand it himself.

Apply this logic to AI prompts: If your prompt enables the AI to produce an answer that "even a layperson can understand," then that prompt is likely high-quality.

How to Do It: A Three-Step Process

Step 1: Ask the AI to Answer as If Explaining to a Beginner

Add the following instruction to the end of your prompt:

Please answer as if you are explaining this to someone who knows nothing about the field. Avoid jargon, and use analogies and examples to illustrate your points.

If the AI’s response becomes vague or full of empty generalizations, it indicates that your original prompt lacks sufficient contextual constraints.

Step 2: Read It Yourself and Ask Three Questions

1. Can I paraphrase this for someone else? If, after reading the AI’s response, you cannot restate the core ideas in your own words, the prompt failed to elicit truly valuable information.
2. Are there any "correct but useless statements"? For example, "pay attention to details" or "stay focused"—these are always true but never helpful.
3. Are there concrete action steps? A good response should include "what to do, how to do it, and to what standard."

Step 3: Iterate on Your Prompt

Based on the issues identified in Step 2, make targeted revisions:

• Response too vague → Add specific scenarios and constraints.
• Contains fluff → Add "Please provide actionable advice; avoid vague generalizations."
• Missing steps → Add "Please list the steps, with each step including specific actions and expected outcomes."

When to Use It (and When Not To)

Suitable Scenarios:
- Writing tutorials, guides, or SOPs (Standard Operating Procedures)
- Explaining technical solutions to non-technical stakeholders
- Preparing training materials or client deliverables
- Using AI as a "teacher" when learning a new domain yourself

Unsuitable Scenarios:
- Highly specialized technical documentation (Feynman-style simplification may sacrifice precision)
- Creative writing (the goal is to "move people," not just "explain clearly")
- Data analysis and code generation (validate with test cases, not with linguistic clarity)

Quick Checklist

• [ ] Added a "explain to a beginner" constraint at the end of the prompt
• [ ] The AI’s response can be paraphrased in your own words
• [ ] No "correct but useless statements" present
• [ ] Includes concrete action steps
• [ ] Iterated on the prompt at least once

Common Pitfalls

1. Satisfied with the first draft: The core of the Feynman Technique is iteration. The first response is almost certainly not good enough.
2. Oversimplification: Explaining to a beginner ≠ talking nonsense. Good simplification retains core logic while removing redundant details.
3. Evaluating without revising: Identifying problems but failing to update your prompt renders the exercise pointless.

One-Sentence Summary

The standard for a good prompt isn’t how professional the AI’s answer sounds, but how clear it is. The Feynman Technique is the ruler that helps you measure that clarity.