A professor at Alcorn State University just pulled off one of the smartest AI-cheating stings we’ve seen. Jason Gibson embedded a hidden instruction in his midterm prompt using white font, telling any AI that read it to sneak the word“Madagascar”into the response. The catch? Students who copied the prompt into a chatbot and submitted the output without rereading it turned in essays about the Industrial Revolution with Madagascar randomly in them. Gibson says 32 of his 35 students across two classes failed that portion of the midterm.
The method is called prompt injection, and it’s invisible to the human eye but instantly readable by AI. Gibson posted the experiment on TikTok, where it racked up 13,000 upvotes, and the story blew up on Reddit’s r/Amazing. Reaction online was mostly unsympathetic, with commenters arguing that not even rereading your own essay is a wild level of careless. Others pushed back, pointing out that students who copied the prompt into a document before writing their own answer could also get caught unfairly, and noting that the numbers haven’t been independently verified.
Love it or hate it, this story is a mirror for how deeply AI has slipped into school and work. The lesson for students is simple: read what you submit, every single time. The bigger question is what this means for classrooms going forward, and whether creative traps are the answer or a sign something needs to change.
Would you have caught the hidden Madagascar instruction, or would your essay have been the plot twist? Tell us in the comments.
About the Author
Local Lawton
Local Lawton is a contributor to LocalBeat, covering local news and community stories.