What Happens When Teachers Use AI to Create Exams: A Personal Story and What Parents Should Know
The first whiff of chalk dust and overheated linoleum still takes me back to that November AP Lit exam. I’d pulled two all-nighters quizzing myself on Blake, Wordsworth, and the dozen Swinburne poems we’d worked through for six weeks, and I had a crumpled study guide sticking out of the pocket of my ripped jeans. When Mr. Hale passed the stapled exam packets down the row, I flipped to the back to see the essay prompts first, like I always did, and my vending machine coffee jitters turned to cold panic. The first prompt asked me to analyze maritime metaphor in Swinburne’s The Sea Mists, a poem I’d never heard of. The second asked for a comparison between two 19th century poems that weren’t on our reading list, weren’t mentioned in lecture, and weren’t in the course anthology we’d all bought at the start of the semester.
I checked my syllabus three times under my desk, flipped through every page of my lecture notes, and came up empty. I’d studied exactly what I was told to study. There was nothing to do but fake my way through the next two hours, stretching vague generalizations about 19th century poetry into two half-baked essays. When grades came out a week later, I got a C-. I’d carried an A average all semester, and that one exam dropped me to a B+, which knocked me down three spots in my class rank. I’d already submitted my early action application to my top college, and the transcript they received had that unplanned B+ on it.
I showed up to Mr. Hale’s classroom after school, butterflies in my stomach, and stammered through my explanation. He pulled out my syllabus, cross-checked the poems I mentioned, and his face fell. He told me he’d been trying out this new AI tool that was supposed to cut down the time he spent making exams from an hour to 10 minutes. He was already coaching debate, the school’s literary magazine, and grading 120 essays a week that semester, so the shortcut felt like a lifesaver. He’d just typed “11th grade 19th century British poetry unit exam” into the box, copied and pasted what it generated, printed it before first period, and never bothered to check which poems it picked. It looked well-written, the questions made sense, so he used it.
It took three weeks for the administration to process the grade change, and I still had to write a short note to the admissions office at my dream school explaining the error. The whole thing ended up fine; I got accepted, and Mr. Hale became one of my favorite teachers, the kind I still send holiday cards to. But that memory has been popping up a lot lately, now that I’m a parent of a 10th grader, and every school newsletter I get talks about how AI will cut teacher workload and improve assessments.
Teachers are overworked, and I get the appeal of AI as a time-saver. This story isn’t a rant against AI in schools. It’s just a reminder that AI doesn’t know the specifics of your class: it doesn’t know which poems you cut from the syllabus to make time for a unit on Black female poets, it doesn’t know that you spent an extra week on Romanticism and pushed the modern poetry section to after finals, it doesn’t know what your students actually spent time learning. It only knows the general topic, and that’s where the mistake happens.
For teachers who use AI to draft assessments, there are two small, concrete steps that prevent this kind of mess. First, after AI generates the exam, spend 10 minutes going through every prompt and question, highlighting every text, term, or concept, and cross-check each one against your actual syllabus and assigned materials. Delete anything that your students didn’t cover. That takes less time than it takes to grade a single essay, and it avoids hours of back-and-forth later. Second, swap out one or two generic AI-generated questions for ones that reference your specific class discussions. For example, instead of a generic question about Wordsworth’s use of nature, use one that references the debate your class had about whether Wordsworth’s views of nature hold up today. It’s a small change that makes the assessment actually test what your students learned, not just generic knowledge AI pulled off the internet.
For parents, if your kid comes home saying an exam had a bunch of untaught material, don’t just tell them to talk to their teacher. Give them a concrete plan. If they notice the problem during the exam, tell them to raise their hand and ask the proctor to write a note on their exam that says multiple prompts cover unassigned material. That creates a record right away, instead of having to argue about it after grades are posted. If the problem comes to light after the exam, help your kid make a short bulleted list that names which questions were off-topic, and tuck a copy of the class reading list or syllabus next to it. That makes it easy for the teacher to see the problem immediately, instead of having to track down all the information themselves. When I talked to Mr. Hale, I just rambled and forgot to bring my syllabus, which added an extra day to the whole process.
Last month, my son’s math teacher used AI to make a midterm review sheet, and it included three problems on material the class hadn’t covered yet. My son brought his syllabus, pointed out the discrepancy, and the teacher fixed the review sheet that afternoon. No harm done. My own bad experience ended up fine too, but it was three weeks of unnecessary stress that could have been avoided with 10 minutes of a teacher’s time. We’re still figuring out how to use AI well in schools, and that’s okay. But it’s worth remembering that AI is just a draft. It still needs a human to check that it fits the kids sitting in the classroom.
Please indicate: Thinking In Educating » What Happens When Teachers Use AI to Create Exams: A Personal Story and What Parents Should Know