Is AI Actually Helping My Child Learn? How to Tell

Your child's homework looks great. The paragraphs are fluent, the answer is right, it was done in half the usual time. And yet something nags: did they actually learn anything, or did the AI just hand it over? It's the right question to ask — and there's a reliable way to answer it.

The trap: a polished answer isn't a learned one

The core risk with AI isn't that it's wrong (though it can be). It's that it's fluent. Reading a clear, confident explanation produces a powerful feeling of understanding — psychologists call it the “fluency illusion” (Koriat & Bjork, 2005) — that often isn't matched by real knowledge. And when a tool does the hard cognitive work for you, you don't get stronger, in exactly the way watching someone else lift weights doesn't build your muscles. Researchers call this cognitive offloading (Risko & Gilbert, 2016): convenient in the moment, costly if it replaces the thinking rather than supporting it.

The good news is that learning leaves a fingerprint, and its absence does too. You don't need to see the screen to tell which happened.

Four signs AI is helping

• They can explain it unaided. The knowledge survives the screen being turned off.
• They make the AI quiz them. Asking to be tested is the behaviour of someone building memory, not borrowing an answer.
• They catch the AI's mistakes. Noticing when it's wrong is only possible if they actually understand the material.
• The work is theirs — imperfect and personal. Real understanding has rough edges and a voice; pure AI output is suspiciously smooth and generic.

Four signs it isn't

• Perfect work they can't explain. The biggest red flag of all.
• Copy-paste rhythm. Question in, answer out, no thinking in between.
• Homework grades up, test grades flat or down. The gap between supervised and unsupervised performance is where offloading shows.
• Defensiveness. “I used AI” offered as the end of the conversation rather than the start of one.

What genuinely helpful AI use looks like

Helpful use has a shape, and it maps neatly onto how learning actually works. Instead of asking AI for the answer, a student uses it to:

• Learn — get a concept explained a different way when the textbook didn't click;
• Verify — be quizzed, marked, and told what they got wrong and why;
• Explore — be given a harder, unfamiliar problem to stretch into.

These are three of the six steps in the Velvet Method, and they share one feature: in every case the student is doing the effortful thinking, and the AI is the coach, not the player. That's the line between a tool that builds a mind and one that quietly replaces it.

If the work suddenly feels easy, be suspicious — real learning is supposed to feel effortful.

That's not a motivational slogan; it's a finding. Cognitive scientists talk about “desirable difficulties” (Bjork & Bjork, 2011) — the struggle of recalling, explaining and applying is the very thing that builds durable memory. A tool that removes all the difficulty often removes the learning with it.

What to do as a parent

Questions parents ask

References & further reading

1. Koriat, A., & Bjork, R. A. (2005). Illusions of competence in monitoring one’s knowledge during study. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31(2), 187–194.
2. Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676–688.
3. Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: creating desirable difficulties to enhance learning. In Psychology and the Real World (pp. 56–64). Worth Publishers.
4. Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning. Psychological Science, 17(3), 249–255.