Why AI Should Be Banned on Every Graded Test
Three weeks ago, I stood in my classroom staring at two essays that were, for all intents and purposes, identical. Not plagiarized—there was no copied text, no matches in any of the plagiarism checkers. But both essays used the same obscure transitional phrase, both had the same slightly formal cadence, and both meticulously dodged any genuine personal voice. My students, two girls who sit on opposite sides of the room and have never once been friends, had produced eerily similar work.
When I asked them about it, both shrugged. One said she “just wrote it.” The other said she “used a tool to help organize her thoughts.”
And there it was. The elephant in the room that we as teachers have been dancing around for the past two years. AI isn’t just showing up in classrooms anymore—it’s sitting in every seat. The question is what we do about it, especially when it comes to testing.
The moment it clicked for me
I teach tenth grade English. I’m not a tech-phobic teacher; I have students type their final essays, I use a class website, and I genuinely think educational technology has a place. But about a year ago, I gave my students an in-class timed essay. They’d been reading Fahrenheit 451, and I asked them a pretty straightforward question about how the novel relates to modern society. The whole thing took forty-five minutes, pen on paper, no devices out.
The results were revealing. Some essays were messy, disjointed, and opinionated. Others showed flashes of insight mixed with sophomore-level grammar mistakes. And you know what? It was the best set of essays I’d read in months. But what mattered more was what the essays didn’t have: that strange, hollow confidence. The polished but empty prose that makes my stomach hurt when I read it.
The contrast was just too stark. The same students who had submitted pristine five-paragraph essays on their take-home assignments couldn’t replicate that quality under test conditions. Some of them panicked. One girl literally started crying because she couldn’t access her “notes” (which turned out to be a chatbot conversation she’d been having all semester).
That day changed how I think about assessment. If I want to know what my students actually know, I can’t compete with an algorithm that has read every book ever written and writes at a doctoral level. And more importantly, I shouldn’t have to.
Why graded tests matter
Let’s be clear about the purpose of a graded test. It’s not about getting a score to fill in a report card. Tests provide a snapshot of what a student can do independently—what sticks when they’re standing alone in the desert of their own knowledge. That’s a deeply uncomfortable experience, but it’s also a clarifying one.
Here’s the thing about education that gets lost in the AI debate: making mistakes is where the learning actually happens. That moment when a student stares at a blank page, panics at the prompt, and then somehow pulls together a partial answer with the fragments they remember? That’s the moment the brain rewires itself. The struggle is the pedagogy. When AI plugs in the answer, the struggle disappears, and so does the learning.
I saw this play out with a student I’ll call Marcus. Marcus is a bright kid, a little lazy, but capable of doing well when the material interests him. Last fall, his essays were consistently getting B-pluses and A-minuses. He’d submit polished, confident analysis of poems I knew he didn’t love. When I finally sat down with him and asked why his essays sounded different than his in-class comments, he stopped pretending. “I just copy what the AI says and then change it around,” he told me. “Why would I do the hard work when this is easier?”
I couldn’t get angry with him. He was giving me an economic argument—if the goal is a grade, why wouldn’t you use the most efficient tool available? But his reasoning revealed a fundamental misperception about what school is for. He thought the goal was the grade. The actual goal was learning to think, which is hard, slow work that cannot be automated.
What we stand to lose
The current debate is usually framed as a binary: ban AI or embrace it. But tests occupy a particular space. Everything in my curriculum is now a question of trust. Homework? I have to assume AI wrote half of it. Group projects? ChatGPT is the silent third group member. Practice problems? The answers arrive pre-solved and screenshot-ready.
I’ve had conversations with teachers who say “just redesign your assessments to be ‘AI-resistant.'” They talk about asking for personal reflections, in-class writing, or creative assignments. And that’s good advice, in the same way rebalancing your diet is good advice. It doesn’t solve the problem; it just makes the problem smaller.
Here’s what I believe: AI use should be banned on all graded tests because graded tests are the last remaining space where we can verify what a student can actually do. That’s not about punishment or technology fear—it’s about data. Education runs on feedback loops. Students write, teachers respond, students revise. If the writing comes from a machine, the feedback loop is broken, and everyone’s just going through the motions.
It’s also about fairness. Some parents have told me they let their kids use AI for homework because “everyone’s doing it.” They’re not wrong. But allowing chatbot assistance on graded tests would formalize a divide between students who have resources and parents who know how to prompt a large language model, and those who don’t. That turns assessment into a measure of home support rather than a measure of learning.
The practical side of teaching without AI
Since that in-class writing revelation, I’ve made some changes. Not dramatic ones, but honest ones. My tests are now entirely in-class and on paper. This isn’t about being luddite, it’s about defining the boundaries of testing. I tell my students on the first day of each semester: “In this class, graded assessments measure you—just you, your own brain, your own knowledge and skills. Homework is for learning and practicing. Tests are for knowing what you know.”
I also build in small, low-stakes writing opportunities throughout the semester, so students get used to writing without their phones on their desks. These are things like three-minute free writes at the start of class responding to a prompt, or “minute essays” at the end summarizing what they learned. This normalizes the discomfort of writing by hand, without critical thinking. Over time, students stop panicking about generating ideas from scratch.
Parents can help here too. If your child is coming home and telling you their school just banned ChatGPT on tests, frame that positively. You could say: “That makes sense, it means your teacher wants to see what you can do, not what the AI can do.” Ask them what they think about it. Turn it into a conversation. These conversations are important because they communicate that learning isn’t a performance for a machine, but a relationship with your own mind.
But I want to be honest—I’m still figuring this out. I don’t know if collaborative learning will be more difficult. I don’t know if some students will develop “test dependence” on AI and struggle when it’s gone. And I suspect that the line between AI as a tool and AI as a crutch will keep blurring in the future.
The uncertain road ahead
Last week, I sat in the staff lounge and heard a colleague arguing that the whole concept of exams was over. “Why memorize facts when we have the internet?” she said. “Why write essays when machines can write better ones?”
There’s a version of that argument that makes sense. But it misses the core purpose of an education. School isn’t about the product; it’s about the person. And the person still needs to learn how to think, struggle, and arrive at honest conclusions. That cannot be outsourced.
I have no idea what classrooms will look like in five years. But I know this: when I hand back those in-class essays and see a student’s genuine surprise at their own clarity, I remember why this work doesn’t change. We don’t protect tests to be rigid. We protect them because they’re the place where we tell students: Your thoughts count. All by themselves.
And that message is worth saving, one test at a time.
Please indicate: Thinking In Educating » Why AI Should Be Banned on Every Graded Test