A Mississippi college professor has gone viral after revealing how a hidden instruction exposed dozens of students who allegedly used Artificial Intelligence to complete a midterm exam.
Jason Gibson, a history professor at Alcorn State University, shared a series of TikTok videos explaining how he discovered widespread AI use among students, per Fox News.
According to Gibson, 32 of his 35 students failed a section of the midterm after submitting responses that appeared to be generated by AI.
The assignment asked students to compare technological advances during the Industrial Revolution with those of today.
Hidden in white text within the prompt was an unusual instruction telling students to include the word "Madagascar" somewhere in their response in a way that made no sense.
Gibson said the unexpected word repeatedly appeared in student submissions, despite having no connection to the topic.
"You'd be amazed at the bizarre usage of Madagascar that I saw in these students' assignments," he said in one of the videos.
He added that students who rely on AI should at least review the generated text before submitting it.
After posting the first video, Gibson addressed criticism over his grading process. He said students were given the opportunity to challenge their grades if they believed they had been marked unfairly.
According to Gibson, only two students appealed. One successfully had her grade changed after explaining that she had viewed the assignment in dark mode, making the hidden white text visible.
She believed the instruction was a legitimate part of the assignment.
Gibson also stressed that his goal was not to embarrass students or celebrate failing grades.
"I get no gratification out of seeing students fail. It's not my intent to see you fail," he said.
In another viral video, Gibson read excerpts from student responses that awkwardly inserted the word "Madagascar" into otherwise standard essays, reinforcing his belief that AI had produced the work.
Although the videos gained widespread attention online, Gibson later said he does not expect to use the same tactic again.
He explained that the experiment came after growing frustrated with reading assignments that lacked originality.
He said AI-generated writing often follows a recognizable pattern and removes the student's unique voice.
In a follow-up video posted Friday, Gibson revealed his class's final grade distribution: eight students earned A's, 14 received B's, 10 finished with C's, two earned D's, and one student received an F.
The discussion comes as colleges nationwide continue debating how to address AI in the classroom.
Some educators argue that detecting misuse remains important, while others believe students should instead be taught how to use AI responsibly as the technology becomes increasingly common in the workplace.