Explain it back: the one-line habit that shows whether you learned anything
When you have your answer, close the chat. Write one line, from memory, in your own words: what do I now understand? If the line comes easily, you learned it. If it does not, you have found the exact thing to ask about next. It takes about thirty seconds and needs nothing but a place to write.
Why reading the answer is not enough
A clear answer feels like understanding. You read it, it makes sense, you move on. The feeling is real, but it is a feeling about the answer, not about what you could say tomorrow without it. Three studies show the gap between a good result and something learned.
- Better work, same learning. In a lab experiment with 117 university students (Fan and others, British Journal of Educational Technology, 2025, peer reviewed), students who revised essays with ChatGPT improved their essay scores the most. Their knowledge gain was no higher than that of the other groups.
- Not able to quote it. In one session of a small MIT Media Lab preprint (54 people, not peer reviewed), 15 of 18 people who wrote with ChatGPT could not correctly quote a sentence from their own essay minutes later. In each of the other two groups, 2 of 18 could not.
- Summaries go in shallow. Across seven experiments with 10,462 people (Melumad and Yun, PNAS Nexus, 2025, peer reviewed), people who learned a topic from an AI summary ended up with shallower knowledge than people who used web search, even when the facts were the same.
None of these studies tested the explain-back habit itself, and we will not pretend they did. They show the problem: an answer in front of you is not the same as an answer in you. The one line is a way to find out which one you have.
How to do it
Close or cover the chat
Switch tabs, turn the phone over, scroll away. If the answer is still on the screen, you will copy it and call it remembering.
Write one line from memory
Start with "So the point is..." or "It works because...". Use your own words. One or two sentences is plenty. Out loud to another person works too.
Compare
Look back at the answer. Did your line hold up? Was a key part missing or wrong?
Turn the gap into the next question
If you could not write the line, say so to the AI: "I tried to explain this back and got stuck at the part about X. Ask me a question that helps me get it."
What a good line looks like
| You asked about | Weak line | Better line |
|---|---|---|
| Why ice floats | "It is about density." | "Water molecules lock into an open pattern when they freeze, so the same water takes up more room and is lighter for its size." |
| A bug in your code | "The loop was wrong." | "I changed the list while looping over it, so it skipped every second item." |
| A contract clause | "It is a notice period thing." | "Either side can end it with 30 days' written notice, and email counts." |
The test is simple. Could someone who did not see the answer learn something from your line? "It is about density" names a topic. The better line says what happens.
When to use it, and when not to
Use it when the aim was to learn, understand or decide: study, a new skill, a problem you will meet again. Skip it for chores. Nobody needs to explain back a train time. If you do it after every single ask, you will stop within a week. After the two or three asks a day that matter, it holds.
Ways to make it stick
- Paper: a notebook next to the keyboard, one line per answer, dated. Reading the week's lines on a Friday shows you what you actually took in.
- Ask the AI to ask you: add "When you have answered, ask me to explain it back in one line" to your question, or to your saved instructions.
- In the think pad: after you ask, the think pad shows the question "What did you learn?" with room for one line, and saves it to a journal on your device. This week's lines are free.
Questions
How do I know if I really understood what ChatGPT told me?
Close the chat and write one line on it from memory. If you can, and the line holds up when you check, you understood it. If you cannot, you recognised the answer but did not take it in.
Is this the same as the Feynman technique?
It is a short version of the same idea: explain something in plain words to find what you do not understand. One line after an answer is small enough to do every day.
What if I got the answer without trying first?
Still do the line. Trying first is better, but explaining back afterwards is the part that tells you whether anything stayed.
Should I type it or say it?
Either. Writing leaves a record you can reread. Saying it to a friend or a classmate gives you someone who can ask "what do you mean?"