Why AI Should Be Banned on All Graded Tests: A Teacher’s View
On a Friday afternoon in late October, I sat with a stack of essays about The Great Gatsby. The sun hit the library table in a way that made the half-readable handwriting seem worse. I had already gotten through ten papers, all with the usual mix of clumsy syntax and half-formed ideas. Then I reached the eleventh.
The opening paragraph was, in a word, polished. Every sentence was parallel. The word “elusive” appeared twice and was used correctly. The thesis felt airtight. And it didn’t sound like the student who had spent the past month asking me whether “yellow symbolizes money” was a strong theme. It didn’t sound like him at all.
When class met Monday, I pulled his essay out and asked the question I’d been turning over all weekend: “Where did you get this?” He didn’t flinch. “I wrote it. Then I asked AI to help me make it clearer.” To him, that seemed obvious. He had put in the work, and the AI had polished it. The grade, after all, should be about the ideas, right?
Maybe, in a perfect world, that would be true. But a graded test isn’t just measuring what you produce. It’s measuring what you can produce on your own, under conditions that require your brain to do the heavy lifting. When a student runs their essay through an AI tool before turning it in, the feedback I give no longer describes what the student actually did. More importantly, the student never sees where their own voice ended and the machine’s began.
I want to be fair here. I’m not one of those teachers who thinks AI is useless for learning. On a low-stakes assignment, I can see using it as a brainstorming partner, or as a tool to help a student see how to build a stronger transition. But graded tests are a different beast. They are, at their core, a way for me to take an honest snapshot of a student’s skills at a given moment. If AI is allowed anywhere near that snapshot, the picture stops being honest.
What convinced me that the ban needs to be absolute—not just a “cite your AI use” policy—was a conversation I had with another student, a junior named Maria. She had used an AI app to check her math homework for months, mostly to see if her logic held up. By the time we had our unit test, she was struggling. “I keep messing up the steps,” she told me. “But when I check with the AI, the steps make sense.” She had developed a dependency without realizing it. On the test, there was no AI, and the steps dissolved.
That’s when I started to understand. For most kids, especially those under sixteen or so, the line between “AI helps me understand” and “AI does my thinking for me” is too blurry to be trusted. Adults can set boundaries. A fourteen-year-old who is scrambling for a good grade will take the path of least resistance. And if we build a system that lets them lean on generative AI for high-stakes work, we are setting them up to never know what they can actually do alone.
So, in my classroom, the rule is simple: AI tools are not allowed on any graded test, quiz, or in-class essay. I tell parents the same thing on curriculum night. It’s not because I hate technology. It’s because I still believe a test should show me what a kid can do without a net.
Putting this into practice has been less dramatic than you’d think. It’s mostly about changing the way assessments are built. First, I have moved more of my formal tests into class time—handwritten, in a room where the only glowing screen is the clock. Students are allowed to bring a one-page index card of notes, but they have to write it out by hand. The act of choosing what to put on the card, of compressing weeks of material into cramped blue-ink letters, is itself a kind of studying. Handwriting also removes the temptation to copy and paste something clever.
For essays, I use a two-step process. The first draft can be done at home, with whatever tools they want. I tell them, quite openly, “If you want to use AI here, go ahead. Use it like a coach.” But when they come to class for the graded essay, they write it in front of me. Usually, I’ll show them the draft they submitted—printed out—and ask them to move their strongest arguments onto a fresh page. They’re not starting from zero. They just have to do the finishing themselves.
I’ve also started doing what I call “talk-back grading.” If a student turns in a piece of writing that seems far beyond their usual voice, I ask them to sit with me for ninety seconds and walk me through their argument—which points confused them, what they cut, why. It’s not an interrogation. It’s a chance to see if the ideas have a heartbeat. Kids appreciate it, once they realize I’m not trying to catch them cheating, but trying to understand how they learn. More than once, a student has admitted, “I don’t actually understand what I wrote.” That admission is worth more to me than a perfect essay.
Now, I know what some people will say. “We should be teaching students to use AI responsibly, not banning it.” There’s truth to that. I do teach it—in small, low-pressure lessons. I show them how to use AI to brainstorm research questions, or how to use it to rephrase one sentence so they can see a different way of saying something. That’s where learning can happen, because the stakes are low and there’s room to mess up.
But on graded tests, AI needs to stay out. Not because students are morally weak, but because a test is one of the only places where we can see what has actually taken root. If we blur that line, we’re not doing kids any favors. We’re just grading a product that neither we nor they truly created.
I don’t have all the answers here. I’ve got colleagues who think I’m being old-fashioned, and others who think I’m not strict enough. The student with the Gatsby essay, when I told him that the grade on the polished version would stand but that he’d have to take the final in class, looked at me with a mix of resentment and relief. He rewrote the essay by hand that week. It was rougher. Some of the transitions were awkward. But it was his—unevenly, imperfectly his.
That essay didn’t earn an A. I gave it a solid B minus, mostly because the introduction was a mess. He probably thought of that as a setback. I saw it as the first time in weeks that I actually caught a glimpse of the writer he’s becoming. For now, that’s enough. I’ll keep the ban in place, keep the in-class handwritten tests, keep the quiet little conversations. And maybe, gradually, students will remember that the point of school isn’t just to get the grade—it’s to leave with a mind that can think clearly on its own.
Please indicate: Thinking In Educating » Why AI Should Be Banned on All Graded Tests: A Teacher’s View