What Happens When You Remove AI Access Midsemester: A High School Teacher’s Real Observation
The coffee on my desk had gone cold by 4:15 last October, and I was halfway through grading rough drafts for 11th grade personal essays when I noticed the pattern. Three of the drafts in my stack were flawless: sharp topic sentences, perfectly structured anecdotes, no crossed-out half-ideas, no margin notes screaming “FIX THIS”. When I pulled Javi’s draft up on my laptop to check the word choice, I recognized the loose, generic lilt immediately. He’d turned in a whole essay written by AI, and he wasn’t the only one.
Javi stayed after that day for extra help, soccer cleats propped against my desk leg, his hoodie still smelling like cut grass from practice. When I pointed out what I’d noticed, he didn’t even lie. “I get home at 8 every night, I have a quiz tomorrow for chem, I just needed it done,” he said, staring at the scuff on my work shoe. “If I turn in something messy, I’ll get a bad grade.” That conversation stuck with me. We’d allowed AI use for brainstorming at the start of the semester, but it had clearly become a crutch for more than half the class. So the next Monday, I announced we were hitting pause: no full AI draft generation for any work for the rest of the semester. We’d adjust for anyone who needed accommodations, but for the most part, all writing had to be theirs from start to finish.
The first week was chaotic. Kids groaned when I passed out blank draft paper. Mia raised her hand halfway through writing time and said, “I don’t even know how to start a paragraph without typing a prompt into ChatGPT.” Javi turned in a rough draft a week later that was half done, with sticky notes stuck to the margins that said “this is bad sorry” and “ran out of time to fix this ending.” The average grade on that rough draft was 12 points lower than the average from the first unit of the semester. I’ll admit, I lay awake that weekend wondering if I’d made a huge mistake.
Two weeks later, though, I started noticing the shift, and it didn’t show up first in grades. It showed up in peer review. Before, when I walked around the room listening to kids talk through each other’s work, most conversations went like this: “It’s good, I don’t have any changes.” Or if they did, it was generic stuff about comma splices. That week, I stopped next to Javi’s table and heard him say, “I don’t know if I should keep the part where my dad bailed on our first hiking trip. It’s too personal, right? But that’s why I started going by myself, so it feels important to leave it in.” His partner leaned in and said, “No, that’s the best part, I’ve had my dad do that same thing, you should keep it.” I’d never heard Javi talk that way about any piece of writing he’d turned in all semester. Before, when I asked him what his essay was about, he’d just shrug and say “what the prompt says.”
It wasn’t just writing, either. We started doing weekly free responses to our To Kill a Mockingbird reading after that, no AI allowed. Before, most responses were generic, regurgitated talking points about Atticus being a moral hero. After the change, kids started writing things like “Atticus makes me mad because he just lets Scout get in trouble for fighting and never asks why she’s upset.” When we had whole class discussions, kids didn’t pause to pull up a pre-written note on their phones anymore. They talked over each other, disagreed, said things that were messy and wrong sometimes, but they were theirs.
Grades caught up a month later. The average on the final personal essay was almost back to where it had been at the start of the semester, but when I pulled every student aside for a 2-minute check-in to ask them to explain their thesis, 78% of them could do it easily. At the start of the semester, that number was 42%. That’s the difference I didn’t expect: it’s not that kids got better grades, it’s that they knew what they had written.
I didn’t end up doing a full ban for everyone, that’s the practical lesson I’ve taken away from this. Lila, who has ADHD and struggles to organize scattered thoughts on a page, still uses AI to make a basic outline for her essays before she starts writing. I just require her to write every paragraph from scratch after that. For teachers thinking of trying this, you don’t have to yank all AI out cold turkey. First, spend 10 minutes asking kids what they actually use it for. Most aren’t trying to cheat; they’re overwhelmed with six classes, sports, part-time jobs, and scared of getting bad grades if they turn in something messy. For parents, you can replicate this at home too: instead of hovering to check if your kid is using AI, ask them once a week “what’s one thing you think about your class work that you haven’t put in your assignment yet?” That encourages them to bring their own voice out, without starting a fight about technology.
It’s December now, and we’re still working this out. Javi got a B+ on that personal essay, which is lower than the A he got on the AI-written essay he turned in at the start of the semester, but he told me last week he taped a copy of it to the inside of his soccer locker. I don’t have all the answers. Next semester, I’m going to try a unit where kids write a draft with AI, then rewrite it themselves, to see how that experiment goes. I know not every kid will thrive with a full pullback, and I’ll keep adjusting rules to fit individual needs. But I do know this: when we removed AI access midsemester, we didn’t just get messier essays. We got more of the kids. And that’s worth the dip in early grades, to me.
Please indicate: Thinking In Educating » What Happens When You Remove AI Access Midsemester: A High School Teacher’s Real Observation