AI Will Answer Your Question. It Won't Tell You Which Question to Ask.

There's a word for the thing AI doesn't have yet. Ingenuity. Not intelligence, not processing power, not the ability to generate a coherent paragraph at three in the morning. Those it has. Ingenuity is something else. It's the capacity to ask the question nobody thought to ask. To look at a problem sideways. To notice what's missing before anyone else realizes something is missing.

And it turns out, that's the part that matters most.

The answer is only as good as the question

Ask AI whether running is bad for you. You'll get an answer about knees. Impact stress. Joint load. Maybe a note about proper footwear. It's not wrong. But it's also not the whole picture.

What about the mental side? The discipline that builds over months of showing up when you don't feel like it. The particular kind of confidence that comes from finishing something hard. The way a long run can restructure a bad week. None of that appears unless you ask for it. And if you didn't know to ask, you might walk away thinking running is a risk to manage, not a practice that changes you.

The answer shaped your understanding. But the question shaped the answer.

This is true of Google searches too, of course. What you type determines what you find. But with AI the effect is amplified. The response comes back so fluent, so confident, so complete-sounding that it's easy to mistake comprehensiveness for correctness. You don't feel the gaps. The prose doesn't leave obvious holes.

The immigration example

Take a politically charged topic. Immigration numbers, say.

You can ask AI for data on how many people have entered a given country over the past decade. You'll get numbers. Charts, maybe. A clean summary. And based on that, you'll form an opinion.

But did you ask about those numbers as a percentage of the total population? Did you ask which sectors of the economy were facing labor shortages at the time? Did you ask about the fiscal contribution of immigrant workers over a ten-year horizon, or the historical patterns of how previous waves of migration played out?

If you didn't ask, you didn't get it. And the answer you did get felt complete. That's the trap. Not that AI lies. It's that a confident, well-structured non-answer is harder to question than an obvious gap.

The same dynamic plays out in software, in medicine, in finance, in any domain where the right framing of a problem is itself a skill.

Where the skill actually comes from

Curiosity gets you part of the way. A curious person asks more questions than an incurious one. But curiosity without experience is a bit like running without knowing how to pace yourself. You burn out on the wrong questions.

The deeper skill is domain knowledge. Not expertise necessarily, but enough exposure to know the shape of what you don't know. Enough experience to sense when an answer is too clean, too simple, suspiciously easy. Enough familiarity with failure to ask "what could go wrong here that I'm not seeing?"

A doctor doesn't just ask "what are this patient's symptoms?" They ask about family history, lifestyle, what else might present similarly. Not because they're more curious than you. Because they've seen enough cases to know which questions the obvious questions leave out.

That pattern-recognition, the ability to notice absence, to read what isn't written, that's what experience gives you. And AI doesn't hand it to you along with the answer.

What this means practically

Using AI well is itself a skill. And it's not evenly distributed.

The people getting the most out of these tools aren't necessarily the most technically sophisticated. They're the ones who come in with enough context to ask the right questions, catch the incomplete answers, and push back when something feels off. They're the ones who know what the answer should look like, roughly, before they ask.

Which means the gap between a good AI user and a great one isn't processing speed or prompt length. It's the knowledge and experience they bring to the conversation before it starts.

AI lowered the barrier to getting answers. It didn't lower the barrier to knowing which answers matter.

That part is still on you.

I write about technology, digital literacy, and occasionally trail running at recepzerk.com.