A history professor turned the tables on AI-assisted cheating by embedding a secret command inside a midterm question, resulting in 32 students being caught using artificial intelligence to answer the assignment. Dr. Jason Gibson of Alcorn State University posted a test question about the Industrial Revolution that contained an invisible prompt designed to trick AI chatbots into including a specific phrase. When dozens of students submitted that exact phrase, the professor had clear evidence of automated cheating.

What You Need to Know

Dr. Jason Gibson added an invisible instruction to a test prompt that told AI chatbots to include a specific phrase in their response. When dozens of students submitted that exact phrase, the professor had clear evidence of automated cheating. The incident highlights the growing struggle educators face as AI tools become more common in classrooms.

How the AI Trap Worked

The trap relied on a quirk of modern language models: they can interpret instructions hidden in white text, small font or invisible formatting that human readers do not notice. Gibson embedded a directive stating that any AI answering the question must include a particular reference about the Industrial Revolution in its response. Students who used tools such as ChatGPT or Copilot to generate their answers copied the AI output verbatim, unaware that the secret trigger phrase would expose them.

  • Invisible instruction: Hidden in the prompt, visible only to AI but not to human readers.
  • Specific trigger phrase: The AI was told to include a unique reference that the professor could easily spot.
  • Verbatum copying: Students submitted the AI-generated response without rewriting or checking the content.

The Scale of AI Misuse in Higher Education

This incident is part of a larger trend. Surveys have shown that a significant percentage of college students admit to using AI for assignments without instructor approval. Many schools have adopted AI detection software, but such tools remain imperfect. The invisible prompt method is a low-tech alternative that requires no specialized software and can be deployed by any instructor willing to craft question carefully. However, the approach raises questions about fairness and whether it inadvertently punishes students who rely on AI for legitimate reasons such as language assistance.

Why This Matters

The cat-and-mouse game between educators and AI users is likely to intensify. As language models become more capable and accessible, detection techniques must evolve. The Alcorn State case demonstrates a clever workaround but also signals a deeper problem: AI tools are now so embedded in student workflows that traditional cheating prevention is no longer sufficient. Institutions must decide whether to invest in better detection, redesign assessment methods or accept AI as a legitimate part of the learning process. The outcome of this debate will affect millions of students and reshape academic standards worldwide.