Alcorn State University professor uses hidden prompt to detect ChatGPT usage

Jason Gibson, a professor at Alcorn State University, successfully identified AI usage among his students by using a hidden prompt technique. He embedded a command in an exam question written in white font, making it invisible to human readers but detectable by AI chatbots when the text was copied and pasted.

The hidden instruction directed the AI to include the word ‘Madagascar’ in its response in a nonsensical manner. Because the exam topic was the Industrial Revolution, the inclusion of the word served as a clear indicator of automated assistance. Out of 35 students, 32 failed the stage of the examination process by falling for the trap, as their submissions contained the irrelevant keyword.