People who are accused of something they didn’t do feel a certain kind of fear. You are now seventeen years old and in the principal’s office. You are told that software has found that your essay was written by a computer. Your paper. The one you stayed on for three nights. This takes place more often than most school administrators want to admit in schools across the country.
Since ChatGPT became famous in late 2022, teachers have been nervous, which is understandable. In less than thirty seconds, the tool can make a good five-paragraph essay. Schools did want a defense, of course. No one stopped to think about whether or not that defense worked, or who it might have hurt in the process.
The idea behind AI detection tools like Turnitin’s now-expanded software is pretty simple: human writing is hard to predict, but AI writing is not. The software checks each word choice to see how “expected” it is and marks text that uses too many likely patterns. On paper, the reasoning makes sense. In real life, it falls apart in ways that are quietly terrible for the students.
The problem is that not all well-organized, clean writing comes from a chatbot. Many schools encourage students to use software like Grammarly to help them with their grammar. This makes students’ writing more predictable. If a student doesn’t speak English as their first language or was taught to write in strict, formal structures, their work may look “AI-like” to a computer that has never been trained to understand them. The output is marked by the software. The student will be punished. The real meaning behind the words is not taken into account at all.
There’s something off about that picture that makes it hard to stay with it. The students who are most likely to be falsely accused are usually the ones who are already at a disadvantage in school. These are the students who need help more, write in a second or third language, or whose education focused on formulas rather than flourish. An algorithm that is set up to find machine patterns ends up punishing differences between people.
What’s worse is that it’s all hard to see. AI detection tools are more like “black boxes” than plagiarism checkers, which show you a matched source—here’s the website, here’s the sentence, you decide for yourself. The score shows up. In the small print, there is a disclaimer that says it “might be” AI. The people who made the tool won’t fully back it up, so teachers and administrators have to use it to make important decisions about academic misconduct.
There are educators who are fighting back. More and more teachers are letting students appeal almost automatically when they raise AI flags, choosing to trust the person over the machine. Even though that gut feeling is right, it can’t be used as a policy. Individual decisions made under stress can’t be used to run a fair school system.
At this point, it’s still not clear if most schools are even taking this seriously. Going back to handwritten tests in class and putting more emphasis on detection software keeps getting picked because it seems like the clear choice. But being sure is not the same as being right.
How to catch students using AI isn’t the real question. That’s why we give chatbots work that can be easily copied and then trust a broken algorithm to tell the difference. It’s not a problem with the technology. It looks like a problem with the curriculum, but it’s not.
Students, on the other hand, have to use a system that is already suspicious of them. There are some of them who are actually cutting corners. A lot of them aren’t. At the moment, the tools used to tell the difference aren’t good enough to make that call—not consistently, not fairly, and not without consequences that can last for years for a teen.
What we’re going through now seems like the beginning of something that will take a long time to work out. That should work, right? It’s not as good for the students who are trying to prove they wrote their own work right now.
