Why AI Should Be Banned on All Graded Tests: Lessons From My Classroom
I was grading essays on a Tuesday evening, a stack of about thirty response papers on the civil rights movement. Normally, I can tell within a few seconds which essays are going to be good, which are going to be average, and which are going to be a struggle to get through. But that night, something felt off.
The first paper was solid. The second paper, also solid. The third one used the same opening phrase as the first. By the time I got to the tenth, I realized three of the essays sounded exactly like each other—identical transitions, identical vocabulary, identical structure. Not plagiarized exactly, not word for word, but clearly formed by the same source. I couldn’t prove it then. But I knew.
A few days later, I had a conversation with Marcus, a student who had always been an average writer. His test essay was clean, polished, and mechanical. He couldn’t tell me what his thesis was when I asked him to summarize it. He fumbled. And when I asked what he meant by “societal paradigm shifts,” he couldn’t explain it. He knew he was caught.
The school didn’t have a formal policy on AI at the time. We had a staff meeting the following week where it came up. Some teachers said AI was just another tool, like a calculator. Others, including me, pushed back. Because a calculator doesn’t produce your thinking. It doesn’t replace the actual cognitive process of solving a math problem. There’s a huge difference between using a tool to support learning and using it to complete the task for you.
That’s the core argument for banning AI on graded tests, not as a punishment but as a way to preserve what tests are actually for.
When I give a graded test, I’m not just checking if a student memorized facts. I’m looking at their thinking process. Can they construct an argument? Can they support it with evidence? Can they express themselves clearly and originally? AI can mimic all of those things convincingly—better than many students can. But the moment I can no longer see the student in the answer, I’ve stopped assessing the student and started assessing a machine that wrote for them.
I think back to a moment in my own parenting life that made me realize how easy it is to fall into this trap. My daughter, who was in seventh grade at the time, asked me to check her homework on the American Revolution. It was a simple assignment—five short answer questions. Nothing too difficult. But she had typed it up, and her answers were just… too clean. They used words like “significantly impacted” and “a pivotal turning point.” I asked her why she had phrased it that way. She said that’s what the internet said when she looked it up.
She hadn’t asked me to help her think through the answers. She had found a way around that.
That’s the problem. When we allow AI to be part of graded work, we’re not just letting students take a shortcut. We’re removing the struggle that makes learning stick. If you look at cognitive science, there’s a massive body of research on the testing effect—the idea that retrieving information from memory strengthens retention. Graded tests, paperwork, and quizzes are all retrieval practice. They’re designed to make kids work through their own knowledge and find out what they genuinely understand.
AI bypasses all of that.
When a student uses AI to answer a test question, they’re not thinking about the material. They’re not wrestling with a problem. They’re not connecting information from different parts of their brain. They’re reading a polished paragraph that another model generated, and they’re copying it. The person who writes the essay gets a grade. But the person who is supposed to be learning gets nothing.
I know the counterarguments. Some people say AI is a fact of life now, and we need to teach students to use it responsibly. We should teach digital literacy. We should integrate AI into assignments. Maybe yes—in project-based work, in design, in research. But graded tests are different. They are the last remaining space where a student must work from their own head. If we give that up, we’re done.
That staff meeting didn’t really resolve anything. We left without a firm policy. I decided to implement one on my own before the next test. When the class came in, I put a sentence on the board: “Use of AI on this exam will result in a zero and a meeting with administration.”
One student raised their hand and asked, “How would you even know?”
That was a fair question. I asked him to tell me about the keyword for the essay question. He couldn’t do it. I said, “That’s how I’d know.”
We agreed on a policy: no phones, no computers, no tablets on test days. Students would write on paper. The first attempt at this policy felt awkward. Some students complained. A few said they didn’t know how to write essays anymore because they’d been drafting everything with grammar tools and AI editors. One student, Emma, wrote a complete essay with one major spelling error and a fragmented conclusion. She apologized for the mess. I told her the mess was the point.
Something began to shift after that. It wasn’t magic and it didn’t fix everything. But the conversations about the material got more real. When students had to write by hand, I saw their actual sentence structure. I saw what they knew, where they got stuck. I could finally see them again.
Parents sometimes worry that banning AI puts their kids at a disadvantage, because other schools don’t have the same rules. My response is that the test isn’t the product. The essay isn’t the product. The learning is the product. If your child can use AI to write a perfect essay but can’t speak about the topic without it, what’s actually been learned?
I still catch one or two students trying to sneak an AI-generated answer in sometimes. They’ll have periods and commas in all the right places that don’t match their usual writing style. It’s a sign of how normalized this has become. They’re not cheating in the traditional sense. To them, AI is just a tool they tapped on their phone without much thought.
The policy at school is still inconsistent. Some room tests remain unproctored. Some teachers haven’t updated their materials. It’s going to take another year or two to figure out something that holds up. But one thing is clear to me now that wasn’t clear before: a graded test is not a place for shortcuts. It’s one of the few spaces in school where we ask a student to trust their own mind.
And that’s worth protecting.
Please indicate: Thinking In Educating » Why AI Should Be Banned on All Graded Tests: Lessons From My Classroom