Why AI detectors can't save integrity
The first thing educators need to accept is that AI-writing detectors do not work reliably. They generate false positives - flagging genuine student writing, disproportionately for non-native English speakers - and false negatives, missing lightly edited AI text. Because a wrong accusation can derail a student's record, decisions with serious consequences can't rest on a probabilistic tool that is often wrong.
Detection is also an arms race the detectors are structurally losing: as models improve and paraphrasing tools proliferate, the gap between AI and human text narrows. Building integrity policy on detection is building on sand.
Why bans don't work either
The other reflex - forbidding AI entirely - is unenforceable at home and increasingly out of step with reality. Students will graduate into workplaces where using AI well is an expected skill, so a blanket ban trains them for a world that no longer exists while doing little to actually stop use.
That doesn't mean 'anything goes'. It means the policy question shifts from 'how do we stop AI?' to 'when and how is AI use appropriate, and how do we assess learning honestly around it?'
Redesigning assessment for the AI era
The most effective response is to change what and how we assess, so that offloading the work to AI no longer produces a good grade or bypasses learning:
- ✓Assess process, not just product: drafts, outlines, revision history and reflections show the thinking AI can't easily fake.
- ✓Use in-class and oral components: presentations, vivas and supervised writing verify understanding directly.
- ✓Set higher-order tasks: analysis, personal connection, local context and critique are harder to fully outsource.
- ✓Make assignments authentic: tie work to a student's own experience, a specific class discussion, or current local events.
Teach responsible use, don't just police it
The most future-proof move is to bring AI into the open. Set clear, assignment-specific expectations - where AI is allowed, where it isn't, and how to disclose it - and teach students to use it as a tutor and drafting aid rather than a ghostwriter. Citing AI use, critiquing its output, and understanding its errors are themselves valuable skills.
This reframes integrity from a cat-and-mouse game into education: students learn to think with AI while still being accountable for their own understanding. That's a lesson that serves them long after the assignment is graded.