What Happens When You Remove AI Access Midsemester? A High School Teacher’s Experience
The groan went up through my 10th grade English classroom the second I finished my announcement mid-October. Javi, who sits in the back row and usually only speaks up to ask if we can watch a movie, slouched into his chair and crossed his arms. “That’s not fair,” he said. “I use it to fix my run-ons. How am I supposed to do that myself?”
I’d spent the first six weeks of the semester letting students use AI tools for writing, following the advice I’d seen all over education forums: frame it as a tool, like spell check or a dictionary, instead of banning it outright. Homework turn-in was 10% higher than it’d been last year at the same point, and most essays were polished, well-structured, and hit all the learning targets I’d set. It looked like it was working, at first.
Then the discussions started feeling off. We were three weeks into The House on Mango Street, and I asked the class why so many of them had written that Esperanza’s shame about her family’s home was the core of the story. Half the class avoided eye contact. Mia, who’d turned in a near-perfect essay on the topic the week before, shrugged and said “It sounded right. AI suggested it.” That’s when I started looking closer at the work. Kids couldn’t explain the jump between their evidence and their thesis. They’d say “that’s what the tool said” when I asked follow-up questions. The hard work of figuring out what you actually think, then putting it into words, was getting done for them. So I made the call: starting week 8, no more AI access for any brainstorming, drafting, or editing. We’d do all rough work in notebooks or on unblocked Google Docs, I showed them how to disable the ChatGPT extension on their school Chromebooks, and I sent a step-by-step guide to parents for blocking AI sites on home wifi and personal phones.
The first two weeks were messy. Homework turn-in dropped 15% overnight. A third of the class turned in half-finished drafts covered in cross-outs and spelling errors. Grades for the first essay without AI were 12 points lower on average than the first six weeks. I got two emails from parents asking if I was behind on curriculum, and one asked if this was some kind of experiment that would hurt their kid’s GPA. I almost backed off the first week, honestly.
But then things started shifting. Three weeks in, we were talking about the scene where Esperanza is sexually assaulted at the carnival, and Javi raised his hand. He hadn’t volunteered a comment in class the whole semester before that. “I wrote that she doesn’t talk about it much ‘cause she thinks it’s her fault for going with the guy,” he said, twisting the edge of his notebook. “I know that’s not super polished, but that’s what I got from it. I didn’t know how to make it sound fancy, so I just left it like that.” No one laughed. A couple other kids nodded, and one said she thought the same thing but hadn’t wanted to say it. That’s the kind of conversation I haven’t had in this class in three years of teaching it.
A week later, Javi’s mom emailed me. She said he used to finish English homework in 10 minutes flat after school, then scroll TikTok. Now he sits at the dinner table rambling about what Esperanza should have done, and complains that writing a thesis is the worst thing he’s ever had to do. It’s the first time, she said, that he’s ever talked about what he’s learning in school outside of report card time.
What I’ve noticed, that I didn’t expect, is that the difference isn’t just in performance—it’s in how kids talk about the material. When they used AI, they talked about “the essay” or “the prompt” like it was something separate from them. Now they say “I think” or “I messed up this part” or “I don’t get this yet.” It’s more honest. For struggling students, that’s a game-changer. I had Lila, who had a 92 average the first six weeks, turn in a 68 on her first AI-free essay. I pulled her aside after class, and she admitted she’d never actually learned how to write a body paragraph—she’d just had AI write the whole thing for her because she was scared of getting a bad grade. We spent one lunch period working through how to connect a quote to her idea, step by step. Her next essay was a 78, and the one after that was an 82. She told me last week, “I still hate writing, but now I know I can do it, instead of just faking it.”
If you’re a teacher or parent thinking of trying this, there’s a few concrete things that made this easier for me. First, don’t ban AI right at the start of the year. Letting kids use it for a few weeks, then removing it midsemester, lets them see the difference for themselves instead of you just lecturing them about cutting corners. Second, make the steps to remove access clear and shared, not a secret punishment: show your kid how to block the sites themselves, so it’s a mutual agreement, not you sneaking around to catch them. Third, expect lower grades and more mess for the first four weeks. Don’t panic and reverse course right away—that mess is where the actual learning happens.
It’s only been five weeks since I removed access, and I don’t have final data or a neat conclusion. Final exams aren’t for another month, and I’m still not sure if I’ll keep this exact rule next semester. Last week, Javi stayed after class and asked if he could use AI to check his grammar once he writes the whole essay himself. I told him yes. We’re still figuring out what works. The biggest change I’ve seen is small: kids don’t sound like they’re reading someone else’s words when they talk about what they’re learning. They sound like themselves. That’s more than any perfect essay or high test score ever gave me.
Please indicate: Thinking In Educating » What Happens When You Remove AI Access Midsemester