AI Detection in Schools: The Wrong Solution to the Wrong Problem

Missouri School Boards’ Association’s Mark Henderson offers commentary on AI detection in schools, and how it’s ther wrong solution to the wrong problem. This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI.

Mike Tyson once said, “Everybody has a plan until they get punched in the face.” For the education world, that punch landed on November 30, 2022. That’s the day ChatGPT was released, and since then, thousands of similar generative artificial intelligence (AI) tools have flooded the market.

Suddenly, students at every grade level had a new tool at their fingertips. It was free. It was easy. It could do a lot of their schoolwork for them. Put yourself in their shoes: why spend time on an assignment you don’t see value in when a chatbot can finish it in seconds?

This isn’t the first time education has been punched in the face by technology. In the 1980s, it was personal calculators. In the 1990s, the internet. In the 2000s, cell phones. But those were minor bruises compared to what generative AI is doing now.

The Detection Problem

Teachers are now expected to detect when students use AI to complete their work, a task they never trained for and that turns out to be nearly impossible. When I work with teachers on this issue, I run a simple activity: I show them four paragraphs, three written by real students and one by ChatGPT, and ask them to identify the AI-generated one. Their success rate is less than 25 percent.

That hasn’t stopped a wave of tech companies from selling AI detection software to schools desperate for a solution. These tools are wildly inconsistent. In one well-known case, a teacher submitted the Declaration of Independence into a popular AI detector and it was flagged as being 97.7 percent AI-generated. Worse, research shows these programs disproportionately flag English language learners and neurodivergent students, penalizing the kids least able to defend themselves against a false accusation.

The damage runs in both directions. Teachers are stressed about proving misconduct they can’t actually prove. And good students are quietly self-censoring. My own daughter admitted to me that she no longer uses “big words” in her writing because she’s afraid a teacher will accuse her of cheating.

The Wrong Fixes

Some schools have gone the other direction entirely, banning technology and requiring students to write by hand in class. Colleges are even bringing back the dreaded Blue Book exam. That’s not a solution either. It just makes everyone’s life harder and leaves students less prepared for a workforce where technology is everywhere. In a world where digital fluency is a baseline job requirement, removing technology from the classroom does students a genuine disservice.

Schools are never going to win this battle, and they need to stop trying to fight it on these terms. AI isn’t going away. There is no reliable way to prove a student did or didn’t use AI in their work. Detection isn’t a solution. Neither is retreat.

The Right Question

The real question educators need to be asking is this: if a student can paste an assignment into ChatGPT and receive a passing result in seconds, was that assignment worth giving in the first place?

That question reframes the entire problem. Academic integrity isn’t primarily a policy problem. It’s a design problem. And design problems have design solutions.

This is the core idea behind what I call the Assignment Makeover Method, a framework I developed after years of working with educators on instructional design. Rather than trying to catch students using AI, the goal is to build assignments that can’t be easily handed off to one. That means anchoring assignments in what students actually need to know and be able to do, and building in elements like personal reflection, demonstrated understanding, and documented process that require a real human to show up and engage with the work.

Consider the difference between asking a student to “write an essay about the causes of World War I” versus asking them to “identify the one cause of World War I most relevant to a conflict happening in the world today, and explain your reasoning in a three-minute class presentation.” ChatGPT can write the first essay in seconds. It cannot replicate the student’s live reasoning, their personal perspective, or their ability to respond to follow-up questions from the class. The second assignment isn’t harder to grade. It’s harder to fake.

This isn’t about making assignments more difficult. It’s about making them more meaningful. Assignments that connect to students’ real lives, require original thinking, or ask students to demonstrate understanding in real time are more engaging, more educationally valuable, and far more resistant to AI shortcuts, all at once.

A Path Forward

None of this is easy, and teachers are already stretched thin. But redesigning assignments doesn’t have to mean starting from scratch. Most assignments can be made significantly more AI-resistant with targeted adjustments: adding a personal connection component, shifting the final deliverable to a live demonstration, or requiring students to document their thinking process along the way.

The education world has adapted to disruptive technology before. Calculators didn’t eliminate the need to teach math. They changed what math teachers needed to emphasize. The internet didn’t make research skills obsolete. It made source evaluation more important than ever. AI will follow the same pattern, but only if educators respond with design rather than detection.

The good news about getting punched in the face? You can recover. And you can learn how to avoid getting hit the same way again. It’s time for education to stop chasing AI and start redesigning around it.