Why AI Use Should Be Banned on All Graded Tests: A Teacher Explains
I didn’t suspect anything at first. The essay in front of me was about the role of nature in Frankenstein, and it was good — really good. Too good, maybe. The sentences flowed in a way that didn’t match the quiet, hesitant voice that had answered questions in class all semester. There was a subtle shift in perspective, a sophistication that felt out of place. I checked the student’s previous work. Short, choppy paragraphs. Simple vocabulary. Nothing like this polished piece. My stomach dropped.
This wasn’t the first time. Over the past year, I’d seen more and more assignments that didn’t sound like the students who submitted them. And it made me think about what we’re actually testing when we give a graded test. We’re not just checking if you know the material. We’re checking if you can think on your own. When a student relies on AI, the test becomes a ghostwriting exercise. The grade doesn’t reflect what they know — it reflects what a language model can produce in thirty seconds.
Later that week, I met with Marcus’s mom. She sat across from me, twisting a napkin. “He worked on that essay for two days,” she said. “He was so proud.” I showed her the essay next to his earlier one. She stared at the pages. “Who wrote this?” she asked quietly. I explained that I’d seen signs — careful grammar, no typos, a rhythm that felt mechanical. “Did he use an AI app?” She nodded slowly. “He said it was just a helper. Like a tutor.”
We sat with that for a moment. That’s the heart of the problem. A tutor explains, guides, checks your work. You still do the thinking. AI does the thinking for you. When kids use it on a graded test — or on a take-home essay that counts as a test — they’re not demonstrating mastery. They’re demonstrating the AI’s mastery. So I’ve made a rule: no AI on any graded assessment. Period.
It’s not that I think AI is evil. I use it for lesson planning, for drafting emails, for brainstorming. But there’s a difference between a tool and a shortcut. Tests exist to show me what you can do without your safety net. If I can’t see your brain working, I can’t know if you’re actually learning. The ban isn’t about punishment. It’s about clarity. I need to know that the “A” I give you is an A for you, not an A for the app you have open on your phone.
For teachers who want to do this, here’s what works for me. First, make most graded work happen in class, where I can see the process. Write on paper, not laptops. For essays, I ask students to turn in handwritten outlines and rough drafts alongside the final version. It takes more time, but it also tells me where the thinking actually happened. Second, I talk openly with students about why. I tell them: “You might get a better grade with AI, but you’ll leave this class without the skills. That’s not a trade I’ll help you make.” Third, for take-home assignments that have to be digital, I require a short oral presentation. They have to explain their thinking aloud, answer a couple of questions. That’s hard to fake.
For parents, the rules are simpler. Have an explicit conversation about when AI is okay in your house. Maybe it’s okay for research, for summarizing a tricky chapter, for brainstorming a topic. But the final answer — the one that gets turned in — has to be their own. Use specific phrases like: “If you can’t explain what you wrote in your own words, then you’re not ready to turn it in.” Also, ask to see their drafts. Not to police them, but to understand how they work. A kid who produces a polished, error-free paper on the first attempt is either a meticulous writer or someone who had a lot of help. Both are possible. But if you never look at their process, you’ll never know.
The day after I announced the rule, my student Kayla stayed after class. “It’s not fair,” she said. “Everyone else uses it. I’m going to get worse grades because I do my own work.” I told her she’d get better grades eventually, because she’d actually learn. She looked skeptical. I get that. For kids, this feels like choosing a slower path. Why should you struggle through a math problem when an app can solve it instantly? Why should you labor over a paragraph when a chatbot can produce a better one? The answer is that struggling through it builds the pathways you’ll need later. Struggling is how you learn. AI can’t do that for you — it can only do the thing for you, which is exactly the problem.
Here’s what I’ve seen in the months since. Students are sometimes frustrated, but they’re also more present. During in-class essays, there’s a quiet focus that wasn’t there before. Pencils move. Pages fill. They ask me questions, and I get to see their actual reasoning. It’s messier. More incomplete. Sometimes painful. But it’s real. One student wrote a rambling paragraph about a novel we’d read, and for the first time all semester, I could hear his voice in the words. It wasn’t elegantly synthesized. It had a grammatical error. But it was his.
I’m still figuring out the balance. AI is not going anywhere, and I don’t think we should pretend it doesn’t exist. Maybe in the future, we’ll have tests that are designed to work with AI — ones that ask you to evaluate the AI’s answer, or to improve it. That could be a valuable skill. But for now, for basic tests that measure whether kids have mastered the material, I’m keeping my ban. I haven’t solved anything. I’m just making sure that when I grade a paper, I know who wrote it. The rest is still uncertainty.
Please indicate: Thinking In Educating » Why AI Use Should Be Banned on All Graded Tests: A Teacher Explains