On this page
An AI explanation can be clear, patient and pitched at your level. The risk is what happens next. You read it, nod, feel that it makes sense, and move on. Feeling that something makes sense is not the same as being able to use it. Here are five quick checks, each taking two to five minutes, that tell you which one you have.
Why does an explanation feel clearer than it is?
Psychologists Rozenblit and Keil named this the "illusion of explanatory depth" (Cognitive Science, 2002). Across a series of studies, people felt they understood how things work with far more precision and depth than they really did. The illusion was much stronger for explanations than for facts, procedures or stories, and most robust where the mechanism is visible. A well-written AI answer is exactly that: a mechanism laid out in front of you.
A second finding makes this worse. In a classic memory study, students who reread a passage did better on a test five minutes later than students who had practised recalling it. But two days and a week later, the recall group remembered substantially more. Rereading also made students more confident (Roediger and Karpicke, 2006). Rereading an AI answer is likely to work the same way, raising confidence more than memory. The fix is to test yourself, not to read again. More on this in why rereading feels like learning.
Five checks to run after any AI explanation
Use a running example. Say you asked why an ice skater spins faster when she pulls her arms in, and the AI explained conservation of angular momentum: with no external torque, L = Iω stays constant, so when the moment of inertia I drops, the angular speed ω rises.
1. Close the tab and explain it back
Close the chat. On paper, explain the idea in four or five sentences as if to a friend in Class 9. Then open the chat and compare. Every gap you find (here, perhaps you forgot the "no external torque" condition) is something you had only recognised, not learned.
2. Change one thing and predict
Ask yourself a "what if" before checking. If the skater halves her moment of inertia, what happens to her angular speed? (It doubles.) What happens to her rotational kinetic energy? Since KE = L²/2I and L is fixed, halving I doubles the kinetic energy too. If that second question surprises you, ask where the extra energy comes from. (She does work pulling her arms in.) This is the step that separates a story from a model you can use.
3. Solve a fresh problem without help
Find one textbook or previous-year problem on the same idea and solve it alone. Only then compare with a solution. If you can explain but not solve, the gap is in setup, not the concept; our post on why you understand theory but fail numericals covers that gap.
4. Find the flaw in a wrong answer
Write down a tempting wrong claim and say exactly why it fails. For example: "Angular momentum is conserved, so kinetic energy is conserved too." If you can name the error (different quantities, and work is done), your understanding has edges, not just a centre.
5. Come back tomorrow
Next day, spend two minutes writing the key idea and one equation from memory, without notes. Because recall after a delay beat rereading in the Roediger and Karpicke study, this is the check closest to what the exam will ask of you.
How to make the AI part of the check
You can turn the tool around. After an explanation, ask: "Give me two questions on this that would expose a common misunderstanding. Do not give answers until I reply." Then answer them in your own words. Our guide to making an AI quiz you properly has prompts that work. Remember that AI answers can contain errors, so when your answer and the AI's disagree, check a textbook rather than assuming either is right.
This is also how we have built Fermi's AI tutor: it checks understanding as it goes and decides whether to explain, ask or show, so the questions come during the lesson rather than after it.
What this looks like in a study session
For a 45-minute session on a new topic, spend about 25 minutes learning (from a book, lecture or AI), 15 minutes on checks 1 to 4, and five minutes writing tomorrow's check 5 prompt on a sticky note. The next time an explanation feels obvious, treat that feeling as a cue to close the tab and try check 1.
