In schools, copying with AI isn't the real problem

Jared Henderson’s video has a title that sums things up: Everyone is cheating (even the professors). Students hand in AI-written homework, teachers use AI to prepare lessons and sometimes to mark. And everyone pretends it isn’t happening.

Why bans don’t work

Many schools’ first reaction was to ban it. But AI is on every student’s phone, free. A ban that can’t be enforced mostly teaches one thing: rules can be dodged.

AI text detectors are unreliable: they flag human-written texts as “generated” (often from people who write very tidily, or aren’t native speakers) and are fooled by a rewrite. Accusing a student based on a detector is unfair. The invisible watermarks some models embed will help, but won’t solve it.

The real problem: what we assess

A take-home essay assessed two things at once: the ability to think (argue, connect, choose) and the ability to produce a text. AI split them apart: producing a decent text now costs nothing. If we keep assessing the final product, we’re assessing who’s best at using ChatGPT.

So the point isn’t “how do we catch cheaters” but “what do we want students to be able to do, and how do we see it”.

What I’d do

1. More process, less product. Assess drafts, choices, revisions. Ask students to hand in their AI conversations too, and assess how they used it: smart questions, checks, critical thinking.

2. More oral assessment. A ten-minute oral says more than a ten-page essay about how much a student understood. Tiring for teachers, but it’s the test AI can’t sit for you.

3. Writing in class. Not everything, but some. Writing without help remains a fundamental skill, like mental arithmetic even though calculators exist.

4. Teaching how to use AI. How to ask questions, check sources, spot a hallucination, when not to use it. Twenty-first-century civics.

5. Using it as a tutor. The same technology that enables copying can be a tireless private tutor: quizzing, explaining in different ways, setting exercises. Students with fewer resources at home could gain most. But only if we learn to use it that way (see the guide on studying with AI).

The real risk

The risk isn’t that students copy. It’s that they grow up without ever doing the struggle that makes learning happen: reading a hard text, getting stuck on a problem, writing an ugly page and improving it. Learning researchers agree: without cognitive effort there is no lasting learning.

AI makes that effort optional. Schools must make it necessary again, and, as far as possible, desirable.

In one line

We don’t need a war on AI in the classroom. We need to change what we assess: fewer perfect texts, more visible thinking.

Sources

Informational content, current as of the date shown. Prices, models and rules change often: always check official sources. Written with the support of AI tools and reviewed by the author.