Why I Now Ban AI on All Graded Tests: A Teacher’s Perspective
I remember the exact moment I knew something had to change. It was during my spring term, and I was reading through a stack of junior American history exams. The prompt asked them to assess the New Deal’s legacy. Most of the essays had that familiar rhythm of rushed thoughts and imperfect paragraphs—clunky transitions, a few awkward phrases, but unmistakably theirs. Then I hit number seven.
James had always been a solid B student. His writing was earnest, but he tended to overuse the phrase “in conclusion” and his historical dates had a tenuous relationship with accuracy. So when I read his exam, my stomach tightened. The prose was polished, almost clinical. It referenced specific economic indicators without a single citation, which made it sound authoritative but strangely detached. It had a vocabulary that hadn’t appeared in a single assignment he’d turned in all year: “ambivalent,” “precedent,” “paradigmatic.” I’ve been teaching long enough to know when a student has had a sudden growth spurt of intellect—and sometimes that happens. But this didn’t feel like growth. It felt like a guest had shown up and taken the test for him.
I called him into my classroom during lunch. The conversation was awkward, the way these always are. James first denied it, then argued that he’d used an AI app to “polish” his rough draft. He’d written the ideas himself, he insisted—ChatGPT just made them sound better. And wasn’t that just like asking a tutor for help? I pointed out that a tutor would ask him questions, challenge his thinking, point out where his evidence was thin. An AI doesn’t care if he understands the New Deal; it just cares if the essay sounds good.
What convinced me wasn’t his admission, but his face when I asked him a simple question. “What did Roosevelt’s ‘second New Deal’ actually change?” I asked. James stared at his shoes. He couldn’t tell me. He’d seen the term in the AI-generated essay but hadn’t learned what it meant. He’d gotten a solid A on a test that was supposed to measure his knowledge of early 20th-century America. In truth, it measured nothing—not even his ability to use AI well. It measured whether he could prompt a machine convincingly.
That afternoon, I rewrote my assessment policy. From then on, AI use is banned on every graded test in my classroom. No exceptions, no “helpful tool” loopholes. Students can use AI for homework, for brainstorming, for constructing outlines. But when it comes to tests, the phone goes in the backpack, and the laptop stays at home. The reason isn’t just about cheating, though that’s part of it. It’s because a test is the one time I can see what’s actually in a student’s head. That’s the whole point of the exercise.
We’ve all seen the headlines about students using AI to write essays or solve equations. Plenty of schools have tried to moderate use, allowing it for certain assignments but not others. But in my experience, that sends a confusing message. If we tell students that AI is a useful tool for learning, we have to draw a line somewhere. For me, that line is a graded test. It’s not because I think AI is evil. It’s because a test without AI is the only honest measure of student progress I have.
Here’s what I tell my students: think of a graded test like a doctor’s visit. When you go for your annual physical, you step on a scale. You’d laugh if the nurse asked, “How much would you like to weigh?” and then adjusted the reading accordingly. A test is a diagnostic tool. If a student gets a B because an AI wrote half their answer, the test tells me the student is ready for a B. But their next teacher won’t know that. When James moves on to AP History, he’ll need to know about the New Deal. The AI won’t be there to whisper in his ear during the AP exam. He’ll sit in a silent room, looking at a blank page, and he’ll have to pull the answer from his own brain. If we never practice that, we’re setting him up to fail at the moment that matters most.
I’m not naive. I know students will always find ways around restrictions. Some will hide AirPods under their hair or use smart watches. That’s why I’ve also changed how I design tests. Gone are the long, open-ended essays that invite AI-generated responses. Now I structure tests with more short-answer questions and in-class writing where I can watch them work. I also have students do a quick oral defense of one answer—just a minute or two, asking them to explain their reasoning. It makes grading slower, but it’s worth it. I get a much clearer picture in that conversation than from a thousand polished paragraphs.
For parents, I know this feels like a murky area. You might be tempted to think that AI skills are the future, and your kid should get comfortable using them. And you’re right—they should. But learning to think without a machine is also part of that future. When my own daughter studies for a test, I make her sit at the kitchen table with flashcards and a pen. I ask her to teach me what she’s learned. If she can’t do it without opening her phone, we know she hasn’t actually learned it yet. That’s not an anti-tech stance; it’s a realism stance.
A few practical things that have worked in my classroom might help at home. First, ask your child to show you the rough work behind an answer. If they can’t produce a messy outline or a few attempted math problems, that’s a red flag. Second, set a simple rule for test prep: no screens for the last hour before studying. It sounds old-school, but it forces the brain to dig for information instead of reaching for a search bar. Third, when your child brings home a test, don’t just look at the grade—ask them to explain one wrong answer to you. That explanation tells you more than any score.
The other day, James came to me after school. He said he’d reworked his essay by hand, without any AI, and wanted me to read it. It was clunkier, less elegant. But every awkward sentence was his own, and each one showed a kernel of understanding. He still had his facts shaky in places. He still wrote “in conclusion” twice. And I was genuinely proud of him.
We have a long way to go, and I’m still figuring out how to handle AI in other areas of my teaching. There are days I wonder if I’m being too rigid, or not rigid enough. But on tests, at least, the door is closed. That feels right for now, and for the Jameses who will come after him.
Please indicate: Thinking In Educating » Why I Now Ban AI on All Graded Tests: A Teacher’s Perspective