The Dinner-Table Question: How to Tell If AI Is Doing Your Student's Thinking

You usually cannot tell whether AI did your student's thinking by looking at the work. You can almost always tell by asking one question.

If you have a teenager and a phone in the house, you have probably had the worry. Your student turns in good work, the grades are fine, and yet you cannot quite tell whether they are learning or just outsourcing. The essay is clean. The answer is right. But did they actually think it, or did a chatbot think it for them?

Here is the good news and the hard news in one sentence: you usually cannot tell by looking at the work, but you can almost always tell by asking one question.

Why "how much time" is the wrong question. Most of us reach for the screen-time frame, because it is the one we know. How long were you on it. How many hours. But time is not what decides this. A student can spend an hour using AI to check their reasoning and come out sharper, or spend two minutes having it write the one paragraph that required a real view, and come out hollow. The clock does not measure the thing you actually care about.

The thing you care about is whether your student kept the deciding. Using AI to brainstorm, to catch a mistake, or to explain a hard concept is fine, and it can genuinely help. The line is not whether they used it. The line is whether they still did the thinking, the checking, and the judging, or handed those over.

The better question, and where to ask it. You do not need to monitor apps or read over their shoulder. You need to move the conversation from "how long were you on it" to "what did it write, and what did you decide." The dinner table is a better tool than any monitoring app, because the tell is not in the document. It is in whether they can stand behind it.

So ask them to explain it. Not to defend themselves, just to talk you through it. "Walk me through how you got there." "Why did you pick that example over the other one?" "What is the part you are least sure about?"

What you are listening for. A student who did the thinking can do this, even clumsily. They can tell you why. They can point to the part they wrestled with. They can disagree with a piece of their own work. A student who handed the thinking over usually cannot. The signs are quiet but consistent:

• They cannot explain their own reasoning in their own words.
• The work arrived without any visible struggle: no false starts, no crossed-out ideas, no "I got stuck here."
• There are no revisions, because there was no draft to revise, only a paste.
• They cannot defend a choice, because they did not make one.

None of these is proof on its own. Together, and over time, they tell you where the tool is sitting.

What to do about it. You are not trying to catch them. You are trying to keep them in the seat where judgment gets built. A few things help, and none of them require you to become the AI police:

• Make "explain it to me" a normal, low-stakes ritual, not an interrogation. The point is the muscle, not the verdict.
• Separate the two good uses of AI, starting and checking, from the one costly use, deciding. It is fine to use it to begin and to catch mistakes. The middle, the actual view, should stay theirs.
• When something feels off, ask, do not accuse. "Show me how you'd do this one without it" teaches more than "did you use AI for this."

The reassuring part. This is not a case for banning the tool or panicking about the future. AI is not the enemy of learning. Used well, it is a real aid. The whole point is smaller and more manageable than the headlines suggest: your job is not to keep the tool away from your student. It is to make sure that when it matters, your student is still the one doing the deciding.

That is the whole of it. Not how long they were on it. Who decided.