My Daughter Was Accused of Using AI—and We Couldn’t Prove She Didn’t
Last Tuesday, my daughter came home from school and threw her backpack against the stair rail before I even got the chance to say hello. “Mr. Aldridge thinks I used ChatGPT on my history essay,” she said. “He showed me a score from some detector. He says I have to see the principal tomorrow.”
I stood there in the hallway, holding a bag of groceries, trying to remember what I knew about AI detectors. Not much. She stood with her arms crossed, waiting for me to fix it. I’d never felt so useless.
Here’s the thing: she wrote that essay. I know because I watched part of it happen. She had a yellow legal pad on the kitchen table three nights before, jotting down facts about the California Gold Rush. She showed me her outline, with arrows and crossed-out words. She complained about her thesis for an hour. But none of that counts as proof.
I asked to see the essay. She pulled it up on her laptop, and I checked the revision history. There were a few small edits, but she had typed most of it in one long sitting the night before the deadline. That’s just the way she works. She always has been a last- minute person. No teacher was going to see that as evidence of anything other than what it was—a rushed, honest effort.
Mr. Aldridge had used one of those detector tools that flags passages as “likely AI.” He printed the report and circled the percentage. He didn’t show her anything else. He just said, “You need to think about academic honesty before tomorrow.” She was fifteen. She is not a passionate kid, she is a quiet one. She came home and shut herself in her room.
I emailed Mr. Aldridge that night. I tried to be calm, not confrontational. I wrote: “I understand concerns about AI, but could you share what specifically in the essay raised the flag?” His reply came back in under an hour. He said the detector identified “syntactic patterns and uniformity of tone” that resembled AI. He included no examples. I asked if the tool provided any reasoning. He said it didn’t. I asked if she could rewrite in front of him. He said that wouldn’t show she had written it originally.
I sat with that for a while. You can’t prove a negative. There is no way to show that something was pulled from your own head. You can describe your sources, your thought process, your false starts—but all of those are just words. They can be invented too. The more I thought, the more I realized the accusation itself had reversed the burden of proof. She was guilty until she could demonstrate she was innocent, and the only demonstration that counted was something that no longer existed.
The next morning, I went with her to the principal’s office. The principal was a decent man, the kind who says “integrity” a lot but means well. He asked her to talk through her research. She did. She mentioned a book from the library and a website she used. He nodded. Then he said, “I understand, but the detection report is what we have to go on.” I asked if the report was considered stronger evidence than a student’s account. He didn’t answer directly. He said we could appeal, but we would need “compelling documentation”—drafts, notes, earlier versions that showed her thinking. She had thrown away the legal pad. She was already halfway through crying.
I’m not writing to paint Mr. Aldridge as a villain or my daughter as a perfect angel. I know plenty of students use AI. I’m not naive. My daughter has even talked about how kids in her grade use it to speed through worksheets. That doesn’t make her one of them. But the system has made it nearly impossible to tell the difference, and that leaves teachers and families stuck.
Here’s what I’ve learned since then. First, those detector tools are not reliable. They have been shown to flag original writing as synthetic, especially if the writer is clear, measured, or uses certain sentence structures. My daughter tends to write in even, straightforward paragraphs. That works against her. Even OpenAI itself has changed its position on its own detector, and yet schools still use them as if they were definitive. It is wild to hold a kid’s grade hostage to a statistical probability.
But I didn’t know that when I first talked to Mr. Aldridge. If I had, I might have asked him to show me how the detector worked. I might have taken a piece of my own writing and run it through the same tool, to demonstrate how it flags humans. A few parents have done exactly that, and it has changed minds. It didn’t change Mr. Aldridge’s, though. He said, “The output confirms my concern.” There is no arguing with that kind of certainty.
Some practical things did emerge from this mess. Another parent in the district told me that she now has her kids do all their writing in Google Docs and never clear the history. She turns on “suggestion mode” so every edit leaves a trace. That sounds logical, but it doesn’t help if a kid writes their first draft by hand or uses a separate app. It just creates a performance on top of the work.
Another approach is to ask teachers to assign more staged writing. I brought that up in a conference with all of my daughter’s teachers. I suggested they break essays into smaller parts: outline, rough draft, peer review notes, final draft. That way there’s a paper trail that’s harder to fake. A few teachers nodded. One said, “That’s a lot of extra grading in an already packed semester.” I understood. I teach myself—I’m at the local community college—and I know what the workload is like. Staged writing takes more time.
What I wish we had done differently: I wish I had asked for an oral defense, on the spot, not as punishment but as a conversation. I wish I had asked my daughter to write a short paragraph in class, by hand, and then asked the teacher to compare it stylistically to the essay. That might have shown enough. But I didn’t think of it in time. We were all just stuck.
We did try one thing. After the meeting, I emailed a friend in the English department at a nearby university. She looked at the essay and said it read like a “solid high-school paper with some awkward transitions.” She couldn’t say it was human, because there is no way to say that about any text. But she did note that the vocabulary was uneven, which was not typical of AI. I sent that to Mr. Aldridge. He didn’t reply.
In the end, the zero stayed. My daughter has stopped bringing it up. She now types her first drafts in a plain text editor, then pastes them into Google Docs, hoping the less uniform style won’t trip a detector. That seems absurd—she’s altering her writing to avoid a false accusation rather than to improve it. But it’s what we’ve done. She still doesn’t want to talk about history.
It’s been a month now. The grade book shows a zero. Appeals require documentation, and we don’t have any. I’ve started keeping every scrap of paper she writes on, just in case. It feels strange to hold onto a crumpled sheet of notes like it’s a legal document. But in this new world, it is.
I don’t have a clean ending. I don’t know if Mr. Aldridge ever doubted his decision. I don’t know if my daughter will trust a teacher again the same way. We are still figuring out how to live in a system where suspicion is weighted so heavily. For now, she writes in one app, copies to another, and leaves her laptop open while she works. That is the best we can do. It isn’t proof of anything. But it’s what we have.
Please indicate: Thinking In Educating » My Daughter Was Accused of Using AI—and We Couldn’t Prove She Didn’t