Why AI Should Be Banned on Graded Tests: A Teacher Explains
The essay came in early, which was unusual for Marcus. He was a decent student, but he had a habit of finishing things in a hurry—often on the bus that morning. This time, the paper was clean, well-organized, and the topic sentence in each paragraph was so perfectly parallel that it looked like a table of contents. When I asked him about his research process, he shrugged and said he had spent a lot of time on it. That was when I noticed the tell: a small, almost invisible dash—the kind you get when you type two hyphens and your device converts them automatically. Marcus didn’t know how to make that dash. He used the kind with spaces on either side, the old school way. The AI didn’t.
I didn’t confront him right away. I just looked at him and thought about how many times I’d seen this in the past few months. A student who struggled to string together three paragraphs suddenly turning in a piece with a subtle argument and a confident conclusion. A math test where every step was perfectly labeled, even though the student could barely explain his reasoning out loud. We all knew what was happening. The question was what to do about it.
Our school district eventually made a decision: no AI on any graded test. That sounds simple, but it’s actually a bigger deal than you might think. For the past two years, we’ve been told to embrace technology, to teach digital literacy, to prepare students for a world where AI is a tool they’ll use at work. That’s all true. But a graded test is something else entirely.
A test is not the same as a homework assignment. Homework can be collaborative; you can look things up, talk to a friend, even ask your parents. Tests are designed to be done alone. The whole point is to see what you know when no one is helping you. When you allow AI into that space, you’re not testing the student anymore. You’re testing the machine. The grade becomes meaningless, and students know it. They’re not stupid. They look at the person next to them who spent three days memorizing formulas and got the same score they got by typing the question into a chatbot, and they start to wonder why they bothered learning anything at all.
I remember a girl named Anna who aced every online quiz last year. She would finish in half the time and get perfect marks. But when we had our first in-class written response, her hands were shaking. She stared at the blank page for ten minutes, then wrote two sentences and gave up. After class, she told me that she hadn’t actually read the novel we studied for three weeks. She said, “I just asked the AI to summarize each chapter for me before the quizzes. I thought that was what we were supposed to do.” She wasn’t trying to cheat. In her mind, she was being efficient.
That’s the heart of the problem. Many kids don’t see AI as a substitute for thinking; they see it as a shortcut to a better grade. But a graded test is one of the few places where shortcuts are exactly the wrong thing. When you’re in a job interview, you can’t pull out your phone and ask it to answer the questions for you. When you’re in an emergency, you can’t ask a language model to tell you what to say. You have to think on your feet. A test is practice for that moment.
So what does a ban actually look like in practice? It’s not about surveillance cameras or locking up phones. For us, it means that any graded test that contributes to a final grade is done in the classroom, on paper, with me in the room. That’s it. No online platforms for those tests. No take-home essays that count for 30 percent of the grade. If it’s graded, you do it in front of a person. That sounds old-fashioned, but it’s actually the simplest solution. I can watch who is looking at the ceiling and who is looking at the screen. I can see the kid who is biting his pencil and the kid who is clearly just copying from somewhere. It’s not perfect, but it’s honest.
The harder part is talking to parents about it. A lot of parents are worried that a ban sets their kids behind, especially in a world that’s increasingly using AI. I get that. But I tell them the same thing I tell my own children: you can’t learn to ride a bike if you always use training wheels. AI can be a training wheel for research, for brainstorming, for editing—but not for answering a question you should be able to answer yourself. I’ve started giving parents a few specific phrases to use at home. When their kid asks if they can use ChatGPT to study for a test, say, “Sure, if you write down what you learned from it and then put it away.” When they ask if they can use it to check their homework, say, “Do twenty practice problems first, then use AI to see which ones you got wrong—but you have to explain why you made the mistake.” These little boundaries make a huge difference.
One of the things that surprised me is how many students actually feel relieved by the ban. Last month, I had a senior tell me that he appreciated the rule. He said, “There’s so much pressure to use AI that it feels like cheating yourself if you don’t. Like you’re wasting time doing it the hard way.” He admitted that he had stopped doing his English homework because he knew the AI would do it for him. “It felt pointless,” he said. “Why would I spend forty minutes on an essay when I can get an A in ten minutes?” When I told him that the point was to get better at writing, not to get an A, he looked confused. We talked for a while, and by the end, he said he actually missed the feeling of finishing something himself. That sense of accomplishment doesn’t exist when you outsource your brain.
I’m not claiming that banning AI on tests solves everything. Students still use it for homework, and I’m not going to police that. It’s also the responsible thing to teach them how to use AI well—how to fact-check it, how to recognize its biases, how to use it as a starting point rather than a final answer. But there needs to be a line. A graded test is that line. It’s the one place where students have to show what they can do alone, without a safety net. If we erase that line, we erase the last remaining proof that they can think for themselves.
What’s uncertain is how long this will last. Things change so fast in education. We might find new ways to integrate AI into assessments, like having students verbally defend their work or explaining their reasoning step by step. We might even get to a point where the skill is knowing how to ask the right questions, rather than knowing the answers. But for now, in my classroom, the rule is simple: if it’s graded, you do it in front of me. No exceptions.
I still think about Marcus sometimes. He came in later that week and sheepishly asked if he could redo the essay. He said, “I didn’t actually learn anything from that. I just wanted the grade.” I let him rewrite it, on paper, in my classroom, with a dictionary and a printer that kept jamming. He struggled through it, crossed things out, started over twice. When he finally handed it in, he looked tired but also a little proud. “That was harder,” he said. “But I think I actually finished it myself.” That, right there, is the point. The ban isn’t about punishing anyone. It’s about making sure that when a student says, “I did this,” they mean it.
Please indicate: Thinking In Educating » Why AI Should Be Banned on Graded Tests: A Teacher Explains