What Happens When You Remove AI Access Midsemester: A High School Teacher’s Real Experience
Last October, I stood at the front of my 10th grade English class, collecting research paper outlines for our To Kill a Mockingbird unit, and I knew something was off. The scratch on my favorite whiteboard marker was still bleeding through the board, the smell of cinnamon sugar cookies from the lunch line down the hall drifted in, and every third outline I picked up had the exact same polished opening line: “Harper Lee’s To Kill a Mockingbird explores themes of racial injustice and moral growth through the lens of childhood in 1930s Alabama.” When I called on Mia, a quiet student who usually turned in thoughtful, messy work, and asked her to expand on the third point in her outline about Mayella Ewell, she stared at her shoes and mumbled, “I don’t know. AI wrote that part. It sounded right.”
My district had encouraged us to frame AI as a “learning tool” that semester, so I’d let students use it for brainstorming and editing up to that point. But after that exchange with Mia, I decided to test a small experiment: for the rest of the semester, no AI allowed for any of my class work. I laid out the rules simply: all brainstorming, outlining, and first drafts would happen in class on paper, with phones locked in the basket by the door. Any take-home writing had to be handwritten first before students typed it up to turn in. I didn’t frame it as a punishment, just a trial to see how it changed how we worked.
The first week was pure chaos. Kids complained nonstop. Javi, who sat in the back and never turned in late work, told me it was unfair because every other teacher in the building let them use AI. Lila, a straight-A student who worried about every assignment, stayed after class crying because she couldn’t figure out how to start her introductory paragraph without AI walking her through it. I got three emails from parents the first weekend asking if I was aware that AI is a skill kids need for college and future jobs. I wrote back saying I was, and that we’d see how it went.
Two weeks in, I started noticing the first shift, and it showed up in discussions before it showed up on assignments. Before the ban, when I asked for thoughts on the book, I’d get the same generic takes you’d pull from any AI summary. No one argued, no one shared a messy opinion. Now, people were fumbling, stopping mid-sentence, admitting they didn’t know, but then saying things that mattered. One day, Tyler raised his hand and said, “I actually think Atticus is kind of a bad dad. He leaves his kids alone all the time, and he just expects them to be fine with all the hate the town is throwing at them.” No one had said anything close to that all semester. Kids started arguing about it, and kept going for 15 minutes after the bell was supposed to ring. No one checked their phone while they waited to talk. That’s when I realized how much we’d lost when we let AI do the thinking for us.
Grades shifted too, not in the way I expected. The first unit test after the ban, average scores dropped 11%. I wasn’t surprised – kids were used to quizzing themselves with AI-generated flashcards that gave them perfect, memorizable answers, and now they had to sort through their own notes to study. But by the time we turned in the final research papers, the average grade was 6% higher than the first paper of the semester, where AI was allowed. More importantly, 80% of the papers had at least one personal connection to the material, compared to just 20% on the first paper. Mia wrote about how she related to Scout’s frustration with Aunt Alexandra telling her to be more ladylike, because her own grandma does the same thing every Thanksgiving. She got a B+, and she told me after class it was the first paper she’d written entirely by herself since middle school.
It’s not all a success story, though. Javi failed his first paper without AI. He told me he’d been letting AI write almost all his work for two years, because he struggles with dyslexia and thought it was the only way he could get passing grades. We now meet twice a week after school, working on writing one paragraph at a time, and he just passed his last quiz with a C. It’s progress, but it’s slow.
What I’ve learned from this experiment translates to both teachers and parents, and none of it requires a full ban if you don’t want that. First, most kids aren’t using AI to cheat on purpose – they’re using it to avoid the uncomfortable part of learning: bad first drafts, confusion, the feeling that you don’t know what you’re doing. That uncomfortable part is where learning sticks. A concrete small step you can try: if you’re a parent, next time your kid has a writing assignment, ask them to write the first half page by hand without their phone in the room. They can fact check and edit after that, but the raw thinking has to come from them first. If you’re a teacher, do all brainstorming and outlining in class on paper, before anyone touches a laptop. That way you see their real thinking, not a polished AI output. Another small shift I use now: when I ask students what they think about a topic, I ask “what part confused you the most?” instead of “what did you learn?” That question can’t be answered by a generic AI output – it forces them to reflect on their own process.
I still get pushback. Some parents say I’m holding kids back by not letting them learn to use AI properly. Some kids still sneak it on phones when I’m not looking. I don’t know if this permanent mid-semester ban is the right long-term solution. I’m still figuring it out. But what I do know is that three months later, kids still stop me in the hall to talk about that Atticus argument we had after class. They don’t do that for the AI-generated essays they wrote at the start of the semester. Right now, that messy, fumbling, personal conversation is enough for me. We’ll keep working through it one paragraph at a time.
Please indicate: Thinking In Educating » What Happens When You Remove AI Access Midsemester: A High School Teacher’s Real Experience