You haven't finished a book in weeks. Yet you've asked an AI about interest rates, history, strange science facts and at least three questions that began with “Hang on, but why?”
So are you learning—or just having an extremely well-informed conversation with a machine?
Both are possible. The difference is who is doing the thinking.
A good question can teach you more than a quick answer
Imagine asking why interest rates are rising. An AI explains. You nod. Conversation over.
Now imagine following up: What happens to savings? To mortgage payments? What if inflation stays high? Who might still lose out?
That second conversation links ideas rather than stacking disconnected facts. Research on self-explanation—putting causes and connections into your own words—suggests that this sort of mental work can help learning. A 2018 meta-analysis found a positive average effect from prompting students to explain what they were learning.
That research did not directly test whether AI conversations replace books. The reasonable takeaway is narrower: asking, comparing and explaining can be useful learning activities, regardless of whether the starting point is a page or a chat window.
“I understood the solution” is not “I can solve it”
Here's the less flattering version of the story.
You ask an AI to solve a problem. It produces a flawless answer. You think, “Great, I've got it.” Tomorrow, without the AI, you stare at an empty page.
Apparently, the AI learned a lot yesterday. Did you?
A 2025 study involving nearly 1,000 high-school mathematics students found a revealing split. Students with unrestricted AI help performed better while practising, but did worse on a later test without AI than students who had never used it. An AI tutor designed to give learning-oriented help largely avoided that drawback.
A maths experiment is not a verdict on reading, adults or every use of AI. It does show why getting an answer and building a skill are different jobs.
What a book still does differently
A chat is brilliant at following your curiosity. A book can take you somewhere you didn't know to ask about. It also asks you to follow another person's argument, often for much longer than a chat reply.
AI has its own advantage: stop at the confusing sentence, ask for another example, challenge a claim, and connect it to something you already know.
Neither format wins automatically. Books offer a route chosen by an author; AI can help you explore side roads. But if the assistant chooses every turn and does every bit of walking, calling it your mental workout is optimistic.
One tiny change makes the conversation more useful
No new study app required. After asking your questions, look away from the answer and try explaining the idea in one or two sentences. Then ask about the part you couldn't explain.
Retrieving information rather than merely rereading it has evidence behind it: a 2021 review of 50 school-based experiments found learning benefits across many settings. That doesn't mean turning every pleasant conversation into an exam. It means giving your own memory a turn.
Reading less doesn't automatically mean thinking less. A lively AI conversation can teach you something. Just don't confuse the machine's eloquence with your own understanding.
Let AI carry the map. You still have to take the walk.
References (3)
- Bisra et al. (2018), Inducing Self-Explanation: a Meta-Analysis: https://doi.org/10.1007/s10648-018-9434-x
- Bastani et al. (2025), Generative AI without guardrails can harm learning: https://doi.org/10.1073/pnas.2422633122
- Agarwal et al. (2021), Retrieval Practice Consistently Benefits Student Learning: https://doi.org/10.1007/s10648-021-09595-9
