Every teacher reading this has felt it. A submission lands in your inbox. The grammar is too clean. The structure too symmetrical. The argument too measured. You don't have to run it through a detector. You already know.
And then you sit with the harder question. What do you actually do about it?
The Rubric Was Built for a World That No Longer Exists
For decades, homework has rested on a quiet assumption. That the act of producing something, an essay, a worksheet, a reflection, was reasonable proof that thinking had occurred. Output equalled effort. Effort equalled learning.
That assumption is now broken. Not bent. Broken. A student with a phone and ten seconds can produce output that meets every surface criterion of the rubric. The rubric will mark it. The gradebook will accept it. And no learning will have occurred.
Why Detection Tools Are Not the Answer
The temptation is to fight the technology with more technology. Run every submission through an AI detector. Catch the cheaters. Restore order.
This will not work, and most teachers already know it. Detection tools have well-documented false positive rates. They penalise neurodivergent students, ESL learners, and anyone whose writing happens to be more structured than their peers. They turn the teacher-student relationship into a low-trust forensic exercise. And the underlying problem, that the assignment itself no longer measures what we want it to measure, remains untouched.
You cannot detect your way out of an assessment design problem.
The Question Homework Was Always Trying to Ask
Strip away the rubric and the deadline and the word count. What is homework actually for?
It was never really about producing the artefact. The artefact was a proxy. A way for a teacher to glimpse the thinking that produced it. To see whether a student understood, struggled, connected ideas, or moved on too quickly.
AI did not break learning. It broke the proxy. And in doing so, it has forced a question that should have been asked twenty years ago. If the artefact is no longer evidence of thinking, what is?
From "Produce" to "Show Your Thinking"
The most useful shift is small but radical. Move the assessment burden from the artefact to the process around it.
Instead of "write me an essay on the causes of World War I," try "draft an essay using whatever tools you want, then submit a five-minute voice note explaining the three editorial decisions you made and why." The essay becomes the starting point. The thinking becomes the assessable thing.
This is harder to fake. It is also harder to mark. But it actually measures what we always claimed to be measuring.
Three Things to Try This Week
1. Add a verbal layer to one assignment. Pick the next piece of homework and require a 90-second voice note where the student explains one decision they made. You will learn more about their thinking from that 90 seconds than from the entire artefact.
2. Make the source visible. Ask students to submit the prompt they used, the version AI gave them, and the changes they made. Treat AI as a tool to be evaluated, not a secret to be hidden. Students learn faster when they have to defend their edits.
3. Replace one rubric criterion with "evidence of revision." Stop rewarding only the final state. Reward the visible movement from first attempt to final version. Movement is what learning actually looks like.
The Opportunity Hidden Inside the Problem
For most of teaching's history, we have been forced to assess what was easy to assess. Output. Volume. Surface correctness. The thing we actually care about, the quality of a student's thinking, has always been harder to capture and so we mostly inferred it.
AI has made the inference impossible. That is not a loss. That is the end of a long bluff. The teachers who will thrive in the next decade are not the ones who learn to detect AI. They are the ones who finally stop using artefacts as a stand-in for thought, and start designing assessments that surface thinking directly.
The homework problem is not really about homework. It is about whether we are willing to redesign the system that AI has just made obsolete.