Three practical ways to find out what your students have actually learned. It's the best essay Maya has written all year, and your first thought is ChatGPT. Maybe you're right. Maybe the topic finally clicked for her. You'll probably never know for sure, and an AI detector won't settle it. But whether she used AI isn't the question that matters most. What matters is whether she understands what she submitted, and that's something you can actually find out. The three strategies below won't prove who used AI. They'll show you who understands the work and who doesn't, which is what you need to know anyway. Use them with every student, not just the ones who make you suspicious: it's fairer, and it gives you a clearer picture of the whole class.

  1. Ask the student to explain one decision Instead of asking, "Did you use AI?", ask a question about the work itself. Why did you choose this argument? What made you reject the alternative? In a math or science class: why this method and not another? Which step was hardest? If you had another hour, what would you change? A student who developed the work, even with AI assistance, will often be able to explain their choices. A student who simply submitted an answer may struggle to. This isn't a perfect test. Nervous students can struggle to explain their thinking, and confident students can repeat explanations they don't fully understand. But a short conversation can reveal more than another round of AI detection. It also doesn't have to eat your week. Two minutes with four or five students per assignment, rotating through the class over a term, adds up quickly. Or ask the same questions in writing: a short paragraph answered in class, without devices.
  2. Look at the process, not just the final answer A finished assignment tells you what was submitted. It rarely tells you how the student got there. Consider two students who submit equally good essays. One starts with a rough idea, struggles with the evidence, revises an argument, and arrives at a stronger conclusion. The other pastes a prompt into an AI tool and submits the result with minimal changes. The final documents might look almost identical. The learning experience wasn't. When possible, build in a checkpoint or two: an outline due a week before the essay, a first attempt at a problem set before the worked solutions, or a short note on how their ideas developed. You don't need to collect every keystroke or turn every assignment into paperwork. This is also a good place to make AI use visible. Ask students to note briefly how they used AI, if at all: what they asked, what they kept, what they changed. Or have them attach the conversation. When disclosure is expected rather than punished, the relationship shifts from suspicion to transparency, and you learn how your students actually work.
  3. Design one task that requires ownership Add a small step that asks students to do something with what they've produced, on their own. • Apply it to something new. Use the same concept on a fresh example or case. In math, solve a variation where one condition changes. • Find the weak spot. Identify the weakest part of the work and revise it. • Critique an AI answer. Give students an AI-generated response to the same prompt and ask what's wrong, missing, or shallow. This one works with AI rather than against it, and it's hard to do well without understanding the material. The goal isn't to make assignments impossible for AI to complete. That's probably a losing battle. The goal is to make students responsible for understanding, evaluating, and using the answers they submit.

The question worth asking

None of these strategies is foolproof. Each one gives you a little more evidence of what a student actually understands, and together they give you a lot. That evidence matters more than it used to. When a polished submission can be produced in minutes, it no longer tells you much on its own. It also means you can stop guessing. Instead of wondering how a piece of work was made, you can see what the student is able to do with it. The question isn't just whether a student produced the right answer. It's whether they can explain it, challenge it, and use what they've learned. That's what we need our assignments to reveal.