Students are outsmarting artificial intelligence detectors with artificial stupidity

At a glance, the essays look fine — coherent, structured, even insightful. But if you look a little closer, you will find something curious: stray typos, oddly casual phrasings, and strangely shallow logic. These are not careless errors. They are carefully planted.

In classrooms across the United States, college students are entering a new kind of academic arms race. The opponent? Artificial intelligence detection software. The weapon? Artificial intelligence itself — tweaked, disguised, and dumbed down.

According to a recent New York Magazine report, many students now use chatbots like ChatGPT to generate essays, and then purposefully alter them to bypass AI detectors. Some add typos. Others rewrite in clumsier language. One student advised her chatbot to "write as a college freshman who is a li'l dumb" — and shared her method on TikTok.

At Stanford, sophomore Eric described a more layered strategy: feeding chatbot output through several different AI systems in sequence to water down the AI signature. Each transformation makes the final product seem more "human," less algorithmic.

Educators are not amused. At the University of Iowa, teaching assistant Sam Williams noticed a dramatic shift in tone between students' first personal essays and their next assignments. Williams told his students plainly: "Do not use AI. But if you do, at least cheat smart." Ironically, his warning may have inspired more strategic cheating.

What troubles many instructors is not just the dishonesty — it is the lost opportunity. "Whenever they encounter a little bit of difficulty, instead of fighting their way through that and growing from it, they retreat to something easier," Williams reflected.