When Dr. Jason Gibson began grading midterm exams for two summer history courses at Alcorn State University, he noticed a pattern that quickly became impossible to ignore.
The assignment asked students to compare the Industrial Revolution with today’s digital age. Many of the responses were polished, grammatically correct and contained a baffling reference to Madagascar.
Not the island nation itself, but sentences like: “Madagascar floats sideways through the afternoon.”
The bizarre phrases had nothing to do with history. They were the result of a trap Gibson had built into the exam.
Hidden in white font within the assignment prompt was the word “Madagascar,” placed where it made no contextual sense. Students reading the prompt normally would never see it. But anyone who copied and pasted the prompt into an artificial intelligence chatbot, and then copied the chatbot’s response back into the assignment without reviewing it, would unknowingly include the hidden word in their answer.
The result, Gibson said, was striking.
“Thirty-two of my 35 students between two classes failed a portion of their midterm because they all used AI to generate their entire response,” he said in a series of TikTok videos explaining what happened. “Apparently they didn’t proofread it.”
The videos quickly spread across social media, drawing millions of views and igniting debate over the role of artificial intelligence in higher education. TODAY.com reached out to Gibson and Alcorn State University but did not immediately receive comment.
Some viewers applauded Gibson’s creativity. Others questioned whether the tactic unfairly penalized students or reflected a broader failure to adapt to rapidly changing technology.
“You’ve taught them an important lesson on integrity, Sir. Well done!” one commenter wrote.
Added another, “Just so you know, that won’t work much longer. The newer models dont have this issue and even tell the user when someone is trying to prompt inject them.”
Gibson says the point was never to embarrass anyone.
After grading the exams, he sent an announcement detailing why so many students had failed that section of the test. He included a screenshot showing the hidden white text and invited anyone who believed they had been graded unfairly to appeal.
Only two students did, according to Gibson.
One student’s grade was changed after she said that using dark mode had made the hidden text invisible on her screen.
“I get no gratification out of seeing students fail,” Gibson said in a follow-up video. “It’s not my intent to see you fail. That’s weird.”
What surprised him most, he said, wasn’t that students had relied on AI.
It was that so few appeared to have read what they submitted.
Several papers included surreal, AI-generated sentences that would have been obvious to anyone who had spent even a few moments proofreading.
One response read, “Madagascar purple bicycle whispers to the ceiling,” Gibson said.
For Gibson, those passages underscored a growing concern among educators: not simply that students are using AI, but that some are outsourcing the learning process altogether. His comments reflect a debate unfolding on campuses across the country as professors rethink assignments that, until recently, could reasonably assume students were completing their own work.
Some instructors have returned to handwritten essays or in-class exams. Others have embraced AI as a learning tool, requiring students to disclose when and how they used it.
“We’re all just out here innovating and implementing, and trying to hold on to some level of academic integrity in the process,” Gibson said. “You don’t know. I don’t know. “Let’s stop acting like we all have the answer because we don’t.”