College prof hides prompt to catch AI cheaters, finds human nature is pretty much as we thought
A history professor at Alcorn State University in Mississippi devised a simple test to catch students submitting AI-generated essay answers, hiding an instruction in white text within an essay prompt that told any chatbot to insert the word "Madagascar" nonsensically into its response. Every one of the 35 students who sat the test fell for it, submitting essays containing bizarre, out-of-place references to Madagascar, proving they had copied the prompt straight into an AI tool without reading or checking the output. The stunt, shared in a viral TikTok video by lecturer Jason Gibson, has drawn attention as a fresh example of the scale of AI-assisted cheating in education and students' apparent failure to even review what they submit.
The case echoes wider concerns about AI's impact on learning, including MIT research showing reduced brain activity in students who used AI to write essays, and Guardian reporting on rising numbers of British students caught cheating with AI. At Brown University, one professor found take-home exam scores inflated to 96 percent thanks to AI use, only for the same class to average just 48.6 percent on a closed-book final. Gibson said he informed his students of the trick and gave them a chance to appeal their grades; only two did so, with just one appeal succeeding.
- Professor hid an AI-trap instruction in an essay prompt
- All 35 students used AI and got caught
- Echoes wider research on AI harming student learning and critical thinking
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters of the professor's approach argue that embedding a hidden instruction is a legitimate and rather clever way to verify suspicions that had already been raised by oddly generic, AI-flavoured essay answers. They see it as fair to students who do their own work, since unchecked AI use undermines the value of everyone's grades and the wider credibility of the qualification. In their view, students were not tricked into wrongdoing but simply given a test that plainly rewarded honesty and exposed those who outsourced their thinking to a chatbot without disclosure.
The case against
Critics of the method argue that a covert trap is a blunt and somewhat adversarial way to address a genuine grey area, given that many students remain unclear on where permitted AI assistance ends and prohibited use begins. They worry that treating this as a 'gotcha' moment shifts attention away from harder questions about how essay assignments and academic policies should adapt to widespread AI availability, and that it risks branding a large share of a class as cheats without first ensuring expectations were unambiguous and consistently taught.