The Growing Tension Around AI in Education and Student Writing

My nephew is a junior in high school. Smart kid, genuinely curious about history, terrible at sitting still long enough to finish a first draft. Last spring he told me, very casually over dinner, that he had started using ChatGPT to write his English essays. Not as a starting point. As the whole thing. He would cut and paste the assignment questions, look through what he got, perhaps tweak a few sentences, and then submit. This was how he had done things, in his reckoning, eleven or twelve times that year. His grades were fine. Nobody noticed. He felt vaguely weird about it but not weird enough to stop.

I did not know what to say to him. I still am not entirely sure I do.

This discussion keeps replaying itself in my head because it encapsulates the spirit of our time in education. This discussion has little to do with cheating, even though it has everything to do with cheating. It has little to do with technology, even though without technology none of it would be possible. It is about a very gradual transformation of the purpose of school, one that is taking place right under our noses and without us there to stop it.

Teachers Did Not Sign Up for This

If you talk to many teachers at present, you will notice a special kind of tiredness. It is not so much burnout but more exhaustion from the realization that there is an unsolvable question that you have been asked to resolve single-handedly while having thirty-two students waiting outside your classroom.

A friend who teaches ninth-grade English told me she spent part of her winter break redesigning every writing assignment she had used for the past six years. Not because she wanted to. Because she realized that most of them could be completed entirely by AI without a student engaging with the reading at all. The essay on symbolism in a novel. The compare-contrast piece on two historical figures. The personal reflection that was vague enough to be written by anyone or anything. She rebuilt them from scratch, making them specific to classroom discussions, to particular moments in the text they had analyzed together, to things only someone actually in the room would know.

That took her weeks. She will most likely have to repeat the process next year as the tools become even more sophisticated. This is how the AI dialogue unfolds in the context of the classroom: not as a grand philosophical discussion of the future of education, but as a very pragmatic struggle to give assignments meaning.

Students Are Not Villains Here

I want to be careful not to frame this as students behaving badly. Some of them are, sure, in the same way students have always found shortcuts when shortcuts were available. But a lot of what is happening feels less like cheating and more like a rational response to a system that has not updated its expectations.

Imagine it from the point of view of a seventeen-year-old kid. They are being told to write an essay on something that they don’t care about, and this is going to determine their GPA, which is going to affect their application to college, which is going to determine the trajectory of the rest of their life, and they’re supposed to do all of that because of the grades. Four minutes, and it can produce a B-plus essay for you.

That does not make it right. But it does suggest that yelling at students about academic integrity is probably not the lever that changes anything. The lever is making the writing itself matter in a way that the generation cannot replicate. And that is a much harder thing to pull.

The Detection Trap

Somewhere in the last eighteen months, a lot of schools landed on AI detection software as the primary answer. Scan the submissions, flag the suspicious ones, and send the flagged ones to review. Straightforward enough in theory.

In practice it has been kind of a mess. These tools have a documented tendency to flag writing by non-native English speakers, by students who write in a clean academic style, by anyone whose sentences happen to follow a structure the detector associates with AI. One widely circulated case involved a student whose original handwritten essay was typed up and submitted digitally, only to be flagged as AI-generated. The student nearly failed the assignment. These are not edge cases anymore.

On the other side, students have figured out quickly that running AI output through a paraphrasing tool, or just editing it manually for twenty minutes, tends to drop the detection score considerably. So what you end up with is a system that punishes students who write clearly and misses students who are actually gaming it. That is not a solution. That is a different problem wearing a solution’s clothes.

What Schools Are Actually Trying to Protect

Underneath all the policy chaos, there is something worth protecting here, and it is worth naming clearly. Writing assignments in school are not primarily about producing documents. They are about making students think in a structured, accountable way. When a student writes an argumentative essay, they have to decide what they believe about something, gather evidence that supports it, anticipate objections, and express all of that in language that another person can follow. That sequence of cognitive events is the point. The essay is just the record of it having happened.

When AI writes the essay, that sequence does not happen. The student has a document. They do not have the experience of having worked through a hard problem in writing. Multiplied across years of schooling, that gap adds up to something real: a student who arrives at adulthood holding credentials that do not fully reflect what they can do on their own.

There is an honest version of this conversation that most institutional responses are still avoiding, which is that some assignments deserve to be retired because they never actually required thinking in the first place. The five-paragraph essay on a generic theme. The summary-response that could be completed without reading the source material carefully. If AI can do the whole thing without breaking a sweat, maybe the assignment was already not working hard enough.

A More Interesting Use of AI in the Classroom

Here is where I think the conversation gets genuinely interesting rather than just depressing. There is a version of AI in education that does not replace student thinking. It responds to it. Instead of generating the essay for you, it reads the essay you wrote and tells you honestly where it is not working. Where the argument goes soft. Where you have stated something without supporting it. Where a reader would get confused and give up.

That kind of AI is doing something closer to what a good writing teacher does, minus the thirty-student classroom and the stack of papers due by Friday. It is treating the student as the author and engaging with their actual thinking rather than substituting for it. The student still has to write. They still have to revise. The AI is just a more available and more patient reader than most students have access to.

That is the premise behind Thanis. The platform is built around the idea that improving your writing means engaging with what you have already written, not handing the task to something else. For students especially, that distinction is not minor. And if you want to see exactly how different that approach is from just opening ChatGPT and typing a prompt, the comparison Thanis makes with ChatGPT is worth a few minutes of your time.

Where This Is All Going

Honestly? I do not know. I do not think anyone does, and I am skeptical of people who sound too confident about it.

What I do think is that the schools and educators who come out of this period well will be the ones who stopped asking, “How do we stop students from using AI?” and started asking, “How do we design learning so that AI use does not hollow it out?” Those are different questions with very different answers. The first one leads to detection arms races and policy battles. The second one leads to rethinking what writing assignments are actually supposed to accomplish, which is a conversation education was probably overdue for anyway.

My nephew, for what it is worth, did tell me he felt like he had not actually learned anything in that English class. He passed it. He got a decent grade. He could not tell you much about what he read or what he argued in any of those essays. He knew that something had gone wrong, even if he had made peace with it.

That seems like the right place to start: with the honest acknowledgment that getting through something and actually learning from it are not always the same thing and that the tools we use have a lot to do with which one happens.