Accused of AI With No Proof: A Teacher’s Honest Look at a New Classroom Problem
The classroom was empty except for the hum of the projector and the faint smell of a student’s forgotten bag of chips. I was about to turn off the lights when I noticed Maya still at her desk, holding her essay paper. She was a quiet junior, not someone who lingered. She waited until the door closed, then said, “Miss, did you think I used ChatGPT on this?” Her voice was steady, but her knuckles were white against the pages.
I didn’t know what to say. I had run her essay through one of those AI detectors because something felt off. The tool flagged it as highly likely to be AI-generated, but I also knew these detectors aren’t actually reliable. They flag non-native speakers, creative writers, sometimes even old textbooks. Still, the suspicion had planted itself. So I asked, “Why do you think that?”
“Because you gave me a weird look when you handed it back,” she said. “And you wrote a comment asking about my sources. But I did write it. I have an outline and drafts. I saved everything. And yet I still can’t prove to you that I didn’t use a machine.”
That last part hit me. She was right. I had nothing solid, only a tool that guesses. And she, on the other hand, had no way to prove a negative. How does any person prove they didn’t do something? You can’t. That gap between suspicion and evidence is becoming the scariest part of teaching right now.
Her paper was about the effect of social media on teenage mental health. She argued that apps are designed to keep you scrolling by triggering little spikes of dopamine, and she cited a few studies. The vocabulary was smarter than what she usually used in her quick writes. But she’d also improved all year. She had gone from awkward sentence fragments to solid paragraphs. So this could have been real growth.
But I was suspicious. I’d even asked a colleague, and they shrugged. “If the detector says it, assume they cheated,” they said. That felt wrong. I was an authoritative adult with a grade sheet, and she was a kid with no defense. I decided to talk to her rather than issue an accusation.
I told her I had been suspicious. I apologized for the look and the comment. She didn’t relax. Her voice got a little louder. “I spent three nights on this. I looked up those studies myself. I organized everything. And now you think a machine wrote it because my grammar was too good? That’s so messed up.”
I could hear the genuine frustration. And I couldn’t argue with her logic. I was using a feeling, a flawed algorithm, and her previous writing struggles as evidence. None of that was concrete. I had no proof, and she had no way to offer proof.
This situation isn’t going away. Every week I read about new AI detection tools, and every few months I hear that those same tools are found to be biased. Administrators send out emails that say “monitor for academic dishonesty,” but they never tell us how. The result is that teachers like me are forced to become detectives armed with hunches.
That week with Maya, I did something that felt small but important. I asked her to walk me through her process. She pulled out her outline—a real one with arrows and crossed-out ideas. She showed me her annotated sources, all printed and highlighted. She showed me a rough draft that had things scratched out in pencil. None of that was airtight proof, but it showed a mind at work. It made me believe her. But I know a student who actually used AI could have faked all of that, too. The difference was that I chose to trust her.
Then she said something I couldn’t shake. “It’s like I’m guilty until proven innocent, but I can’t prove I’m innocent. So what was the point of that?”
I told her I wasn’t accusing her. I told her to keep up the good work. She thanked me, but the wariness didn’t leave her eyes. And for the rest of the semester, I noticed she cut back on raising her hand. Maybe she felt distrust, or maybe she was a shy junior. I’ll never know.
Since that day, I’ve made changes to my assignments. Not because I think all my students are cheating, but because I want to avoid putting them in that impossible position. For every major paper, I now add a short in-class writing piece. It’s nothing dramatic—a warm-up asking them to summarize their argument or explain one editing choice. That gives me a sample of their actual voice without pressure. I also ask them to submit their brainstorming notes and any rough drafts along with the final version. I don’t grade those extra pieces, but they help me see how a piece of writing evolved.
I also changed the way I talk about AI. Instead of announcing “I’m checking everything,” I try to be honest. I say something like, “I want to see that you’re thinking. If you use any tool, even an AI one, that’s okay if we talk about it. But you have to show me how you used it.” That seems to lower the anxiety. When I do feel concerned about a piece, I don’t rely on a detector anymore. I go straight to the student and say, “This sentence doesn’t sound like your usual voice. Can you tell me about what you meant here?” That opens a conversation instead of an accusation.
I still worry about academic integrity. I’m not naive. There are students who use AI lazily and then get away with it. That bothers me, too. But the problem is that suspicion without evidence chokes the relationships that make real learning possible. If a student feels accused, they’ll shut down. And very few students can actually prove their own authorship—they don’t record themselves writing, and why would they? So the burden has to be lighter.
Maya got a B that semester. We never brought up the AI detection thing again. At the graduation ceremony, she waved at me, and I waved back. I like to think she didn’t hate me for that afternoon. But I’m not sure, and I accept that. Learning to live with uncertainty is part of the job now. The best I can do is keep creating assignments that reward process, keep having conversations instead of delivering verdicts, and remember that behind every suspicious page, there is a student with a story. Sometimes that story is the truth. My job is to leave enough room for it to be told.
Please indicate: Thinking In Educating » Accused of AI With No Proof: A Teacher’s Honest Look at a New Classroom Problem