News

At the start of the 25-26 school year, my school introduced new guidelines establishing the role of artificial intelligence (AI) in the classroom. Administrators emphasized “responsible use,” teacher discretion, and clear expectations. But silently among these policy changes, another tool emerged: AI detectors—systems that analyze vocabulary patterns, sentence structure, and word predictability to determine whether a text was written partially or completely by AI. This software is often treated as a definitive way of determining authorship despite its well-documented limitations and ethical concerns. However, when I put the Preamble of the U.S. Constitution—a document written in 1787—into a well-regarded detector called GPTZero, the system flagged it as 12% AI-generated. If a historic and undeniably human-written text can be misclassified, what does that say about the reliability of these programs?

This concern becomes more urgent when we consider how widespread AI detectors have become. Their use has grown significantly between 2023 and 2025, coinciding with the increased popularity of tools like ChatGPT. Schools nationwide, including my own, now rely on GPTZero, Originality.ai, and Turnitin’s AI Writing Indicator to identify violations of academic integrity. On the surface, these systems appear to be functional, and though vendor websites claim 98-99% accuracy, independent studies, and my personal experience, have suggested otherwise. In fact, a Washington Post investigation found that Turnitin’s AI detector actually produced a 62.5% inaccuracy rate.

At the same time, the spread of AI detectors has become not only a common practice—routinely used by teachers to review homework and essays—but also a profitable business for a projected multibillion-dollar industry, selling “trust” as a service. While academic institutions pay for the illusion of control, students bear the consequences: false accusations, grade penalties, and even loss of scholarships. Given that these inaccuracies remain unresolved and often unacknowledged, how can we be certain that students’ best interests aren’t being affected by the growing financial incentives behind such tools?

My experience has led me to develop an unfortunate answer. Over the past few years, I’ve been falsely flagged three times for AI use despite having written each text entirely myself. As a result, I’ve had to rewrite assignments with the oversight of my teachers—a mentally exhausting process. While I’m honored that my writing appears “too perfect,” the repeated questioning undermines my effort, integrity, and, ultimately, my trust in the education system. What’s saddening is that each time my writing is flagged as AI-generated, I will change the way I write, not because I cheated, but because I’m afraid of being accused. This has shown me how these programs, despite their intentions, can work against the very students they’re meant to support.

However, it’s important to recognize that these concerns impact more than just students. Steven Mintz, a history professor at the University of Texas at Austin, notes that his own published work, written long before AI tools existed, was marked as “AI-generated.” Out of curiosity, Mintz conducted further tests, and the results were staggering. George Orwell’s 1936 essay, “Shooting an Elephant,” was flagged as 53.97% AI-generated. And unless Abraham Lincoln discovered the key to time travel, there’s no explanation for how his Gettysburg Address—written in 1863—clocked in at a whopping 100% AI-generated. In contrast, two responses created entirely by ChatGPT were labeled as only 24.67% and 17.86% AI-generated.

The theme seems clear: AI detectors create a paradox. These systems flag texts based on a few key factors. For instance, terms like transformative, underscore, pivotal, or indicate are often flagged. Certain grammatical choices—em dashes, colons, and long sentences—can also influence a detector’s results, meaning the very skills schools encourage are the ones most likely to raise suspicion. This has led some students to intentionally use AI to sound less “AI-like”—a process that ultimately dumbs down their writing. The system is fundamentally backwards. Human writing is flagged, while full or modified AI texts often pass unnoticed.

So how do educators handle students’ overwhelming use of AI in the ever-changing digital landscape? Some claim AI detectors are imperative to filtering what’s genuine versus artificial. But this sentiment isn’t shared by all teachers. Other faculty use alternatives such as Draftback (a real-time revision playback software), document history, in-class writing, outlines, and direct conversations with students. Yet despite these student-centered methods, schools continue to deal with the reality that many students use AI, leaving teachers in a difficult position.

Still, AI detectors are not the answer. Alternatives provide verifiable evidence of authorship and strengthen student-teacher trust. Detectors, on the other hand, offer unreliable probability scores and do students a disservice in the long term. The fact that better options already exist makes the reliance on flawed algorithms unnecessary and, when accusations occur, inexcusable.

At a school like mine, where rigorous courses and academic excellence are a standard, advanced writing is viewed as an expectation rather than an exception. In fact, 40% of students who attend perform in the top 10% nationally. Failing to account for these factors can have an impact extending far beyond a single assignment. If schools want to maintain fairness, its policies must acknowledge the reality that strong writing is not uncommon.

That leads us to a crossroads. Institutions strive to commit minds to inquiry and hearts to compassion, but using AI detectors contradicts both ideals. There’s no “inquiry” when AI replaces our thinking, and no “compassion” in accusing us based solely on a calculation. We shouldn’t be punished for writing well, and our teachers shouldn’t feel pushed to rely on a systematically flawed tool. At the end of the day, this is about human connection—the relationship between the student who writes and the teacher who reads. That connection depends on trust, and trust cannot come from an insentient algorithm. My teachers want us to think for ourselves and take ownership of our ideas. We want our efforts to be recognized and our voice to be heard. Both sides benefit when there’s dialogue—not fear, not suspicion. Because the moment a teacher doubts genuine writing, it’s not just the work being questioned—it’s the student behind it. And we deserve better than that.