When AI-Generated Exams Hurt Student Grades: What I Learned As A Former Student And Current Parent
I’m sitting in my 10th grade AP Literature final, the fan above the blackboard humming so loud I can barely hear the kid next to me breathing, linoleum still sticky from a spilled chocolate milk at lunch. I’ve spent three weeks quizzing myself on symbolism, written three practice essays, and I’m sure I’m going to walk out of this with an A that keeps my early action application to my state’s honors college on track. I flip the last page to the essay prompt, and the first line makes me blink: Analyze the thematic role of the green light in To Kill a Mockingbird.
I know I studied. I know the green light is Gatsby’s. The light at the end of the Radley porch is the one from To Kill a Mockingbird. I raise my hand, walk up to Mr. Hale’s desk, where he’s grading midterms between monitoring the exam, coffee steaming in that chipped World’s Best Dad mug he brings every day. “The prompt says the green light is in TKAM,” I say. He doesn’t look up from his paper. “That’s what’s on the exam. Just answer it.”
I go back to my desk. My eraser, which I’ve chewed half to bits from stress, crumbles when I try to cross out my first wrong note. I’d worked all year to keep that A. I skipped hanging out with my friends on weekends to study, re-read Gatsby twice, stayed after class once a week to go over essay structure with Mr. Hale. That one wrong prompt almost erased all of that. I spend the next 40 minutes stretching to connect the green light to Atticus’s moral standing, tying it to the idea of lost innocence, even though I know it doesn’t belong anywhere in that book. I leave the exam knowing I blew it.
Five days later, I get my grade back: 62. A D. That drops my overall course grade from an A to a B+, and my school’s weighted GPA system would have knocked me down a full tier for honors college consideration. I cried in the bathroom after class, then went to my mom when I got home.
My mom didn’t shoot off an angry email to the principal that night. She sat with me at the kitchen table while I ate a pop-tart, pulled out my printed study guide, and highlighted the line that Mr. Hale had written in bold at the top: “Final essay prompt will ask you to analyze a key symbol from The Great Gatsby.” She helped me write a short, clear note explaining the mismatch, then we made an appointment to talk to the department head the next day. It turned out Mr. Hale had a toddler with a stomach bug that week, he’d generated the entire exam with AI the night before it was due, printed it, and never read a single question. He’d just trusted that the AI got it right. The department head corrected my grade by the end of the week, and I kept my A.
This was almost 15 years ago, before AI was the everyday classroom tool it is now—back then, it was just a weird new resource teachers were testing out, but now I see it everywhere. I work as a part-time English tutor for high schoolers, and I have a 9th grader of my own, so I see how many overworked teachers are turning to AI to crank out lesson plans and exams when they’re already juggling 150 kids, parent meetings, and extracurriculars. I don’t blame them. I get it. But that day has stuck with me, because it’s not a one-off mistake I laugh at now—it’s a common risk that’s easy to miss, and there are simple, low-effort ways to avoid it.
First, for teachers: If you’re using AI to generate any part of an exam, set aside 10 minutes before you hit print to read every question out loud and cross-check it against your study guide and assigned texts. That’s not a big ask. Last month I used AI to generate 20 practice multiple choice questions for a student prepping for the AP Lit exam, and it mixed up the protagonist of Jane Eyre with Rebecca. I caught it in two minutes of skimming, but if I’d just sent it to my student, we would have wasted an hour of study time. Second, when a student flags a question that seems off, don’t assume they’re just confused or trying to get extra points. Tell them you’ll check it after the exam if you can’t look right then, don’t just brush them off. That small step would have saved me a week of stress back in 10th grade. Third, if you’re grading an exam and a student’s answer sounds totally off base, take 30 seconds to check the prompt against your unit material before you mark it wrong. Half the time, the mistake isn’t the student’s—it’s the AI’s.
For parents: If your kid comes home with a terrible grade that feels totally out of character, don’t start by yelling at the teacher. First, ask them to pull out their exam and the study guide the teacher provided. Have them point out exactly what didn’t line up, write that down in a clear, short email, and ask the teacher to take a second look. If the teacher won’t reconsider, ask to talk to the department head, just like my mom did. She didn’t call Mr. Hale lazy or irresponsible, she just said “We know you’ve had a tough week, and it looks like there was a simple error in the exam prompt that affected my daughter’s score.” That was enough to get the conversation started without putting anyone on the defensive.
A couple of weeks ago, my 9th grader came home from biology and told me he’d flagged a wrong question on his unit test—AI-generated, his teacher admitted, that mixed up the definitions of mitosis and meiosis. His teacher fixed the grades for everyone that afternoon, because she checked the question right away when kids pointed it out. That’s how it’s supposed to work. AI cuts down on work for overstretched teachers, which is a good thing, but it doesn’t replace that 10 minutes of human checking. I still see stories every month of kids getting grades knocked down for errors that aren’t their fault, and I know it’s going to keep happening as more teachers use these tools. It doesn’t have to, though. Most of the time, the damage is totally avoidable with just a little bit of extra time up front.
Please indicate: Thinking In Educating » When AI-Generated Exams Hurt Student Grades: What I Learned As A Former Student And Current Parent