Professor Jason Gibson's recent TikTok video has received over ten million views, revealing a shocking and amusing situation: out of 35 students in his two classes, 32 failed the midterm exam because they all used AI-generated answers without even checking them.
Gibson's anti-cheating method was clever and elegant. He embedded a hidden prompt in the exam with white text that is invisible to the naked eye. If students simply copied the questions into an AI tool, this hidden instruction would cause the AI to generate a meaningless sentence containing "Madagascar." When grading the papers, the sudden appearance of the word "Madagascar" in all 32 answer sheets exposed the cheating immediately.
After the exam, Gibson publicly explained the method of screening for cheating and provided an opportunity for appeals. In the end, only 2 out of the 35 students chose to appeal. Gibson shared in his video the absurd "Madagascar" sentences generated by AI, which sparked a heated discussion on social media about AI and academic integrity.
This incident has hit upon a common anxiety facing global education systems: as AI tools become easily accessible, the traditional exam format's ability to prevent cheating is rapidly becoming ineffective. Students no longer even need to read the AI output before submitting it - in Gibson's case, if anyone had glanced at those strange paragraphs about Madagascar, the entire class wouldn't have collectively failed. Perhaps the more urgent question is not "how to stop students from using AI," but rather: when the basic willingness to check AI answers is disappearing, how much of the educational foundation remains?
