Students who prepare to teach what they are learning remember it better than students who study only for themselves, and the gap is not small. The protégé effect explains why, and a 2026 study adds a twist: the effect may not work when the student is an AI. Here is what the research actually shows, how strong the effect is, and how to use it.

What the Protégé Effect Is
The protégé effect is the finding that people learn material more thoroughly when they prepare to teach it to someone else than when they study it only for themselves. The term was coined by Catherine Chase, Doris Chin, Marily Oppezzo, and Daniel Schwartz in 2009, who wrote that students “make greater effort to learn for their TAs than they do for themselves.” The effect does not depend on being a good teacher. It depends on knowing, while you study, that someone else will rely on your explanation.
That last part is the whole trick. When you study for a test, you can skim past the parts you half understand and hope they do not come up. When you study to teach, you cannot. A confused student will ask about exactly the part you skipped, and you know it while you are still reading. So you study differently: you organize more carefully, you slow down at the hard parts, and you go hunting for the gaps, because you have to anticipate where a reader will get confused. The social pressure does the work that willpower usually fails to do.
The effect is easy to confuse with a famous cousin, and the difference matters. The Feynman technique asks you to explain a concept in simple words, usually to an imaginary child or a blank page, and the benefit comes from reformulating knowledge. The protégé effect adds a real audience. It is not just about reformulating knowledge. It is about reformulating knowledge for someone else, and knowing that someone else is waiting for it.
Where the Effect Comes From
The name comes from a 2009 study in which eighth grade students worked with a computer character called Betty. Chase and her colleagues gave the students a program where Betty was a cartoon pupil, and the students taught her about river ecosystems by building concept maps she could reason from. The students who taught Betty read more carefully, revised more often, and learned more than the students who learned the same material for themselves. Lower achieving students gained the most, which is the opposite of what you might expect from a technique that sounds like extra work.
The idea is older than the name. In 1980, John Bargh and Yaacov Schul found that people told they would later teach a passage to someone else remembered it better than people told they would be tested on it, even though nobody in the first group actually taught anything. The expectation alone changed how they took the material in. John Nestojko and colleagues at Washington University in St. Louis extended the finding in 2014. Participants who merely expected to teach recalled more, and tellingly, recalled it in a more organised way, reproducing the main ideas before the details. Expecting to explain had changed how they studied.
There is a third piece of the story, and it is the one that makes the effect practical rather than just interesting. Logan Fiorella and Richard Mayer showed in 2013 that students who actually delivered a short recorded lesson outperformed those who had only expected to teach, when tested a week later. Expecting to teach helps. Actually teaching helps more. The two stack, and the article below says how much each one is worth.
How Strong Is the Effect, Really
Pooled across dozens of experiments, the gain is moderate, and it depends heavily on knowing in advance that you will teach. A recent evidence review puts the pooled gain at roughly g = 0.17 to 0.56, depending on which synthesis you read. Keiichi Kobayashi’s 2019 synthesis in Japanese Psychological Research pooled 28 studies and found an effect size of g = 0.35 for preparing to teach, and g = 0.56 for preparing and then actually teaching. In plain terms, that is a real, noticeable improvement, not a miracle. It is the kind of gain that moves a grade, not the kind that turns a beginner into an expert overnight.
Kobayashi’s 2024 meta-analysis in Educational Psychology Review pooled 39 studies and reported g = 0.27 overall, and it found the moderator that matters more than any other. Teaching after studying with a prior expectation of teaching gave g = 0.48. Teaching after studying without that expectation gave g = -0.02, statistically indistinguishable from nothing. Deciding to explain something after you have already studied it does not produce the effect, because the effect happens during the studying, not after it. The decision has to come first.
How you teach matters too, almost as much as whether you teach. Kobayashi found that teaching a person face to face after preparing produced an effect of g = 0.84, while explaining into a recording with no advance warning produced g = 0.23. A listener who can interrupt, ask questions, and push back forces you to understand the material at a level a silent recording never demands. That is the finding a 2026 study tested against AI, as described in a 2026 guide to the protégé effect and how teaching changes learning.
Why Teaching Changes How You Learn
Three mechanisms sit underneath the effect, and each is worth understanding on its own, because each one tells you how to use it better.
- Organisation: Reading lets you drift through material in a line, one sentence after another. Preparing to teach forces you to ask what comes first, what depends on what, and which parts are load bearing. You have to build a structure, and building the structure is itself the learning.
- Retrieval: Explaining is a retrieval act. Decades of memory research show that pulling information out of your head strengthens it far more than putting it in again. Every time you explain an idea without looking at the page, you are practising recall, and recall is what makes knowledge last.
- Gap hunting: When you know someone else will rely on your explanation, you look for the holes before they do. The part you glossed over, the term you cannot define, the step you cannot justify: these are the places a student will stop you. Finding them while you study is the point, and it is the step most people skip when they study alone.
Notice what is missing from that list: talent, charisma, or teaching skill. None of the three mechanisms requires you to be a good explainer. They require you to prepare as if someone is depending on you, and then to explain as if someone is listening. The pressure is the mechanism. That is good news, because pressure is free, and it is also the reason the effect is so easy to fake badly, by explaining to nobody and calling it teaching.
The 2026 AI Twist: Teaching a Machine
The newest question about the protégé effect is whether the student has to be human. In 2026, Xu, Zhang, Tang, and Lee randomized 96 participants across four conditions: explaining to an LLM playing a tutee, a peer, a challenger, or a near-passive control. On the post-test, no condition beat the control. On current evidence, explaining to an AI tutee does not reliably produce the learning advantage, even though the AI can answer, ask questions, and play the role of a student.
That result fits the older evidence rather than contradicting it. The face to face number, g = 0.84, was measured with human partners, and the recording number, g = 0.23, was measured with no listener at all. An AI listener sits somewhere between the two, polite and responsive but not genuinely depending on you, and the 2026 result suggests that in-between is not enough. A real student creates a real obligation. You know they will be confused if you are vague, and you know they will notice if you dodge the hard part. An AI will not be confused, and it will not notice. It will just keep going.
The honest reading is narrower than the headline. One study with 96 participants is not a verdict, and AI tutors are improving fast. But the direction of the evidence is consistent: the protégé effect is a social effect, not a talking effect, and the social part is doing most of the work. If you want the gain, teach a person, as explained in a 2026 guide to the protégé effect for students and parents. If no person is available, the next best thing is a listener who can genuinely interrupt you, not a screen that nods along.
How to Use the Protégé Effect
Using the effect well takes a little setup, and the setup is where most people go wrong. The most common mistake is studying first and deciding to explain later. That order does not work, because the benefit happens during the studying. Decide who you will teach before you open the book, and study with that person in mind.
Pick a real listener, someone who will actually listen and ask questions. A study partner, a colleague, a friend, a child, a classmate. Tell them you will explain the topic to them, and set a time. The appointment is what creates the expectation, and the expectation is what changes how you study. Then study as if the explanation is the test, because for this purpose, it is. Organize the material into an order a beginner could follow, and mark every place where your own understanding feels thin. Those marks are your study list.
When you teach, do it out loud, without notes, and let the listener interrupt. The generation effect shows the same principle from another angle: producing an answer yourself beats reading one someone else wrote. When your listener asks a question you cannot answer, do not dodge it. Write it down, go back to the source, and explain it again. That loop, explain, get stuck, repair, explain again, is where the deepest learning happens. Three rounds of it will teach you more than three rereads of the same chapter.
Finally, keep the sessions short and the topics small. Explaining one concept well beats explaining a whole chapter badly. A ten minute explanation of a single idea, done properly, will do more for your memory than an hour of vague coverage. And if you have no one to teach, write the explanation as if you do: a short note, in plain words, addressed to a named person who will read it tomorrow. It is not as strong as a live listener, but it is far stronger than rereading.
- The protégé effect is the finding that people learn material more thoroughly when they prepare to teach it to someone else than when they study it only for themselves. The name comes from Chase and colleagues in 2009.
- The effect depends on knowing in advance that you will teach. Kobayashi’s 2024 meta-analysis found g = 0.48 with prior expectation, and g = -0.02 without it. The decision to teach has to come before the studying, not after it.
- Three mechanisms drive the effect: organisation, retrieval, and gap hunting. Preparing to teach forces you to structure the material, pull it from memory, and find the holes before a student does.
- How you teach matters. Teaching a person face to face after preparing gave g = 0.84, while explaining into a recording with no advance warning gave g = 0.23. A listener who can interrupt is worth far more than a silent audience.
- A 2026 study found no learning advantage for explaining to an AI tutee over a near-passive control. The protégé effect is a social effect, and the social part is doing most of the work.
- To use it, pick a real listener and set a time before you study, study as if the explanation is the test, teach out loud without notes, and repair every question you cannot answer.
Frequently Asked Questions
What is the protégé effect?
The protégé effect is the finding that people learn material more thoroughly when they prepare to teach it to someone else than when they study it only for themselves. The term was coined by Catherine Chase, Doris Chin, Marily Oppezzo, and Daniel Schwartz in 2009, in a study where eighth grade students taught a computer character called Betty about river ecosystems and learned more than students who studied for themselves.
Does the protégé effect actually work?
Yes, with a real but moderate gain, and only under the right conditions. Kobayashi’s 2024 meta-analysis of 39 studies reported an overall effect of g = 0.27, rising to g = 0.48 when students knew in advance they would teach, and to g = 0.84 when they taught a person face to face after preparing. Without prior expectation, the effect disappeared. It is a reliable technique, not a miracle.
Is the protégé effect the same as the Feynman technique?
No, although they are related. The Feynman technique asks you to explain a concept in simple words, usually to an imaginary listener or a blank page, and the benefit comes from reformulating knowledge. The protégé effect adds a real audience: the benefit comes from knowing someone else will rely on your explanation, which changes how you study in the first place. The two pair well together, but they are not the same thing.
Does teaching an AI count as teaching?
On current evidence, not reliably. A 2026 study by Xu, Zhang, Tang, and Lee randomized 96 participants to explain material to an LLM tutee, a peer, a challenger, or a near-passive control, and found no learning advantage for the AI tutee condition. A listener who can genuinely interrupt and depend on you matters more than a screen that plays the role. One study is not a final verdict, but the direction of the evidence is consistent.
Final Thoughts
The protégé effect is one of the cheapest learning tools that exists: it costs nothing, it needs no app, and it works because it borrows someone else’s need to know. The catch is that it only works when the audience is real and the decision to teach comes first. Study for a listener, teach out loud, and let the interruptions do their work.
Call to Action
If this guide helped you see studying differently, share it with someone who is learning something hard right now. Explore the Aziz Publishing Knowledge Library for more guides on psychology, including how focused practice builds mastery and how to break a problem down to its basics.