What Happens When You Remove AI Access Midsemester: A High School Teacher’s Real Experience
Mia leaned against my classroom counter after 3rd period last month, crumpling her half-written Scarlet Letter draft into a ball in her fist. “I don’t even know how to do this without it,” she said, nodding at the Chromebook tucked under her arm. She wasn’t wrong. I’d given her and every other 10th grader in my class permission to use AI for the first half of the semester, so this moment was entirely my doing.
I started this year with the common take that banning AI was pointless. Kids were going to use it anyway, so I might as well be upfront about it. I told them it could brainstorm topics, fix wonky grammar, and explain confusing reading passages, just like a peer tutor or a parent helping with homework. For two months, that worked fine on paper. Grades were consistent, turn-in rates were high, no one got caught cheating. But I started noticing a weird quiet in class. When I’d ask an open-ended question like “What do you think really motivated Dimmesdale to stay in Boston?” the same three kids would answer. Everyone else would stare at their desks. When I pulled kids aside to talk about their essays, they’d shrug and say “That’s just what I wrote” when I asked them to expand on their thesis. It felt like they weren’t actually connected to the work.
Three weeks ago, two months into the semester, I sent out a class announcement: No more AI for any class work, from reading responses to midterm essays. Any use would count as plagiarism, same as copying a friend’s homework. I expected pushback, and I got it. The first week, half my class had late work. Kids emailed me at 9pm saying their brains “forgot how to think” without it. Javi, a quiet kid who usually turns in everything on time, stayed after class to tell me he’d used AI to outline every single reading response the first two months, and now he just stared at a blank Google Doc for 20 minutes before giving up. I told him that was normal, and that we’d work through it.
What happened next surprised me. It didn’t happen overnight, but after the first messy week, the whole dynamic of the class shifted. Last week, we were discussing the ending of The Scarlet Letter, and Mia — the same girl who crumpled her draft — raised her hand. Before the ban, she never spoke up. Her essays always had that polished, generic AI tone that said nothing about her. But this time, she said, “I think it’s stupid that she came back and put the scarlet letter back on. If I’d survived all that, I’d move to Europe and never think about Boston again.” The room exploded. Half the class agreed with her. The other half started arguing that she missed the whole point of the book. We spent 22 minutes talking, when I usually budget 5 for end-of-book discussion. No one pulled out a pre-written answer from anywhere. It was just messy, loud, real thinking.
Grades told a similar messy story. The first set of reading responses after the ban had an average score 12 points lower than the set before. Kids turned in work with messy structure, off-topic tangents, grammar mistakes that I hadn’t seen in years. But by the second set, the average climbed back up. When I graded midterm essays last weekend, the overall average was exactly the same as the first unit average from before the ban. But the essays were nothing like the old ones. Half of them had arguments I’d never read in 7 years of teaching this book. One kid argued Dimmesdale was more scared of being rejected by the town’s kids than he was of God. Another said the scarlet letter was just a plot device Hawthorne used because he didn’t know how to write real female characters. None of these are the neat, standard answers AI spits out. They’re just what those kids actually thought.
This isn’t a story about AI being bad for kids, by the way. I still think it can be a useful tool for kids who know how to use it right. But what I’ve learned is that most high schoolers don’t learn to use it as a tool first — they use it as a crutch, because it’s faster and it means they don’t have to risk being wrong. If you’re a teacher thinking of trying this, don’t yank it cold turkey on a major paper. Start with low-stakes work, like 10-point reading responses, so kids have space to mess up without tanking their grade. If you’re a parent wondering if your kid is relying on AI too much for homework, you don’t have to snoop through their computer. Just ask them one question after they finish an assignment: “What’s the part you still don’t get, or the part you disagreed with?” If they stare at you blankly or can’t answer, that’s a good sign they’re letting someone else do the thinking for them. A simple fix I share is to have them say their answer out loud to you before they type anything. It doesn’t take more than 5 minutes, and it gets their own thoughts out before they go looking for someone else’s.
I still don’t have all the answers. Some kids in my class still haven’t adjusted. Javi still turns in half his work late, and he told me last week he still has to close ChatGPT three times before he starts writing every assignment. Grades are still roughly the same, so no one’s getting a massive boost from this change on paper. But when I walk down the hall after class and see a group of my kids still arguing about whether Hester Prynne made the wrong choice, that’s the difference that matters. It’s not about getting better grades. It’s about getting kids to care what they think, even when it’s messy. I’m still experimenting, still tweaking the rules for next semester. But so far, this experiment has been worth the messy first month.
Please indicate: Thinking In Educating » What Happens When You Remove AI Access Midsemester: A High School Teacher’s Real Experience