Why asking matters more than answers in the AI era

For most of history, education was built around a scarce resource — information — and generative AI has made that resource abundant, which means the scarce resource has quietly become something else: the quality of the question. That single shift is why aitutors.me is built around Learn by Asking rather than around delivering explanations, and it's worth spelling out exactly why, because it isn't obvious until you look at what schools were actually optimising for.

What education used to optimise for

A lecture is an efficient way to get one explanation to thirty people at once. A textbook is an efficient way to store an explanation so it doesn't need repeating. A search engine is an efficient way to retrieve one of millions of stored explanations in under a second. Every one of these tools solves the same problem: information used to be hard to get, so getting it needed to be made cheap.

That problem is essentially solved. A GCSE Chemistry student today can retrieve a correct explanation of ionic bonding faster than they can find their exercise book. The bottleneck has moved.

What changed with generative AI specifically

Search engines already made information fast to retrieve — but you still had to know roughly what to search for, and the answer came back as someone else's finished page, not tailored to your specific confusion. Generative AI removes both of those frictions. You can describe your exact, half-formed confusion in your own words and get a fluent, plausible, personalised answer back in seconds, with no need to already know the right search term.

That's a genuine leap, and it's also exactly why the Answer Trap is a new and specific danger rather than an old one with a new name. It's never been easier to get something that looks like understanding without doing the work that produces real understanding.

The scarce resource flipped

Here's the actual argument, laid out plainly:

A student who asks a chatbot "what's the answer to question 4" gets back exactly what any other student asking the identical question would get: a generic, average, unindividuated response. There is nothing distinctive about that transaction and nothing learned from doing it twice. A student who instead asks "I think the answer involves squaring both sides, but I'm not sure why that's allowed here — is it?" gets something no other student's identical homework would produce, because the question carries their specific confusion inside it.

This is not an anti-AI position

It would be easy to read this as "don't use AI." That's not the claim, and it isn't aitutors.me's position either — the whole product is an AI tutor. The claim is narrower and more useful: the value of an AI interaction is determined by the question that goes in, not the answer that comes out. Two students can use the identical tool and get wildly different amounts of learning from it, purely based on what they asked and what they did after.

That's the whole logic behind Ask-Try-Ask as a working cycle, and it's worth reading if you want the practical version of this argument rather than the theoretical one — see Learn by Asking vs just asking ChatGPT for a direct comparison of what each approach actually produces.

The skill that doesn't depreciate

Facts date. The periodic group numbering a student memorises this term will still be true in five years, but plenty of what's memorised for a KS3 exam won't matter at all by Year 12, let alone university or work. Tools change even faster — whatever AI interface exists in five years will look nothing like today's.

What doesn't depreciate is the underlying skill: noticing the exact shape of your own confusion and turning it into a precise question. That skill works on a chatbot, a textbook, a tutor, a friend, or a research paper, and it works whether the tool available in ten years' time is a chatbot at all. This is also why prompt literacy is not the same thing as question literacy — phrasing a request cleverly is a narrow, tool-specific skill; knowing what's actually worth asking is the durable one underneath it.

What this means for how a child should use AI at home

Practically, it means the measure of a good study session with an AI tool isn't "did I get the answer" — that's now trivially true of almost any question, almost instantly. The measure is "did I ask something that made the answer teach me something." A parent watching over a child's shoulder can apply that test directly: is the next thing typed a fresh question, or is the tab about to close?

For the mechanics of what that looks like in each subject, see the Ask Ladder, which breaks question quality into five concrete, checkable levels rather than leaving it as a vague virtue.

FAQ

Hasn't information always been available in libraries and textbooks?

Available, yes — but not instant or free of effort. Finding the right page, or the right person to ask, took time and often a trip. AI collapsed that cost to almost zero, which is the actual change: not that answers exist, but that producing one now takes no skill and no time at all.

Is this argument against using AI for homework?

No. It's an argument for what to do with the answer once you have it. A learner who gets an AI's output and then interrogates it — asks why, asks what if, asks where else — gets far more out of the same three seconds of AI time than one who copies and moves on.

What skill actually doesn't depreciate?

The ability to notice your own gap and ask a precise question about it. Facts get forgotten and tools change, but knowing what's worth asking transfers to every subject, every tool, and every year of school after this one.