We’ve all been there. You highlight the textbook, re-read your notes, and nod along because the concepts seem familiar. You feel ready for the exam or the presentation. But the moment you try to explain the topic without your notes, your mind goes blank. You stumble over words, realize you don't actually understand the logic, and frustration sets in.
This common struggle is called the "illusion of competence." Recognizing terms isn't the same as understanding them. The traditional way we use AI in education—asking it to explain things to us—often reinforces this passive learning style. We nod at the AI's perfect explanation, but we haven't done the heavy mental lifting ourselves.
There’s a better way to use these tools. Instead of asking AI to be your tutor, you should force it to be your student. By combining the Feynman Technique with specific AI study prompts, you can flip the script. You teach the AI, and in doing so, you trick your brain into learning faster and deeper.
The Science: Why Teaching Beats Reading
Before we dive into the prompts, it's helpful to understand why this works. When you prepare to teach someone else, your brain organizes information differently. This is known as the Protégé Effect. Recent studies suggest that treating AI as a "novice student" rather than an expert triggers this psychological phenomenon, significantly increasing the cognitive effort you put into the material seen in Protégé Effect research.
The difference in results is stark. Passive review methods, like re-reading or highlighting, often result in retention rates dropping to just 10-15% after one week. In contrast, active recall methods—specifically learning by teaching—can boost retention rates to approximately 80% in the same timeframe comparing active recall vs passive reading.
The Feynman Technique exploits this by forcing you to simplify complex concepts. If you can't explain it simply, you don't understand it well enough yet.
Step 1: Set the Persona for Feynman Technique AI (The Simulated Novice)
The biggest mistake students make with Feynman technique AI prompts is letting the AI remain too smart. If the AI acts like a professor, it will "fill in the blanks" for you, guessing what you mean even if your explanation is vague. That defeats the purpose.
You need to engineer a "Simulated Novice"—an inquisitive, slightly annoying student who doesn't know any jargon. Research shows that adopting a specific persona, like a curious 12-year-old, forces the user to deconstruct complex vocabulary into plain language.
Try this prompt:
"Act as a curious, intelligent 12-year-old student. I am your teacher, and I am going to explain [Topic] to you. Your goal is to learn from me. Do not access your internal database to fill in gaps. Only rely on what I tell you. If I use a big word you don't know, stop me and ask what it means."
Step 2: The Explanation Phase
Now, type out your explanation of the concept. Don't worry about being polished; worry about being clear. Use analogies. If you are explaining "Gradient Descent" in machine learning, don't talk about derivatives yet. Talk about walking down a misty mountain where you can only feel the slope under your feet.
This phase triggers generative processing. Because you can't rely on the AI to do the thinking, your brain has to synthesize the information, reorganize it, and output it in a new format supporting generative processing.
Step 3: The Ruthless Critique
Here is where AI shines, but only if you command it correctly. Large Language Models (LLMs) suffer from a quirk called "sycophancy"—they want to be helpful and polite, so they tend to agree with you even when you're wrong as noted in sycophancy studies. A polite AI tutor will say, "Great job!" even if your logic is shaky.
You need to break this sycophancy. You need an AI that cares more about truth than your feelings. You must explicitly instruct the model to identify logic gaps.
Use this follow-up prompt:
"Review my explanation above. Do not be polite. Analyze it for logical gaps, undefined jargon, or circular reasoning. If I skipped a step, tell me exactly where you got lost. Rate my explanation's clarity on a scale of 1-10 and tell me what is missing to make it a 10." Feynman technique prompt templates
Step 4: The Teach-Back Loop
Once you’ve received the critique and refined your explanation, use the "Teach-Back" method to verify your success. Ask the AI to summarize the topic, but with a strict constraint: it can only use the information you provided.
The verification prompt:
"Based ONLY on the explanation I just gave you—and ignoring your own outside knowledge—summarize [Topic] back to me. If my explanation was incomplete, your summary should reflect those holes."
If the AI's summary sounds broken or confusing, that’s not a failure of the AI; it’s a reflection of your current understanding. That is your cue to go back to the source material, fill the gap, and try again.
Advanced Tips for Power Users
Once you are comfortable with the basic flow, you can layer in advanced techniques to make your study sessions even more efficient.
- Use Voice Mode: If you are using a mobile app with voice capabilities, do this exercise orally. Speaking often reveals hesitation and knowledge gaps more effectively than writing, as you can't edit your words as easily as you type Oxford AI case studies.
- System Prompts: If your tool allows for "Custom Instructions" or "System Prompts," set the AI to always behave as a naive learner. This saves you from having to paste the setup prompt every time you start a new chat AI system prompt guide.
- The "Why?" Game: Instruct the AI to ask "Why?" after every third sentence you write. This mimics the recursive questioning of a child and forces you to drill down to first principles.
Why This Matters
In an era where information is cheap, understanding is the premium asset. It is easy to ask ChatGPT to write an essay for you, but that outsources the thinking process. By flipping the dynamic and using AI study prompts to teach the machine, you reclaim the cognitive work.
You move from a passive consumer of content to an active architect of your own knowledge. The next time you sit down to study, don't ask the AI for the answer. Tell the AI, "Sit down, I'm going to teach you something," and watch how much faster you master the material.