AI Cheating Isn't a Technology Problem. It's an Assessment Problem.

This week, Denmark's government told upper-secondary schools to start doing three things: require students to verbally defend any written exam they completed at home, monitor student screens during in-school written exams, and push more written work back into supervised classrooms instead of letting it happen at home unwatched. Education minister Magnus Heunicke didn't dress it up: “we have a problem with AI cheating in upper secondary schools.” It's a striking set of measures, screen monitoring software in a school exam hall wasn't on anyone's bingo card five years ago, but the more interesting question for parents isn't whether Denmark is right to act. It's why a written essay ended up being the thing that broke first.

What Denmark actually announced

The rules apply to Denmark's roughly 9,000 upper-secondary students on the two-year Higher Preparatory Examination track, and take effect immediately as the new school year starts. Any written exam completed at home now requires a verbal defense, a student has to explain and answer questions about work they claim as their own. In-school written exams get computer monitoring. And schools are being pushed to move more written assessment on-site, under supervision, rather than trusting it happens unaided at home. The government is also preparing a wider national AI strategy for schools. None of this is a ban on AI. It's an admission that the current way of checking whether a student actually learned something has quietly stopped working.

Why the essay was the weak point

A take-home written essay has always relied on a simple, unspoken assumption: if a student hands in coherent, well-argued writing, they must have done the thinking behind it. That assumption held up fine for decades because producing convincing writing was genuinely hard without doing the underlying work. It doesn't hold up anymore. Generating fluent, plausible-sounding text is exactly what large language models are best at, and nobody watches an essay get written the way a teacher watches a student work through a problem in class. Once the deliverable and the evidence of learning are the same document, and that document is the single easiest thing in the world for AI to produce convincingly, the format itself is the vulnerability. Monitoring software and oral defenses are a real response, but they're a patch on a task that was never built to survive this moment.

• A written essay: the final text is the entire deliverable, and the thinking behind it is invisible
• A working product: the deliverable can be run, questioned, and changed on the spot
• An essay only proves someone can produce convincing text
• A product a student can explain, defend, and modify live proves they understand what they built

What actually resists this, without needing monitoring software

The honest answer isn't “ban AI” or “watch harder.” It's changing what counts as proof of learning in the first place. A real, working product, one that solves an actual problem for an actual user, is a fundamentally different kind of evidence than a paragraph of text. You can ask a student to change a feature in front of you. You can ask why they made a specific decision and watch whether the answer holds up. You can't fake having built something the way you can fake having written about it, because the proof isn't a document, it's a thing that either works or doesn't, built by someone who either understands it or doesn't. That's not a new idea we invented in response to this news; it's the same reasoning behind why every PsyntaxLabs Academy student builds a real product with a real user instead of handing in a written assignment, and why cohorts are capped at six so a founder-mentor is actually present for how the thing got built, not just grading what arrived at the end.

“You can't out-monitor a bad assessment. The fix isn't watching students harder, it's asking them to prove something a shortcut can't fake.”

PsyntaxLabs Academy

The question worth asking about your child's schoolwork

You don't need to wait for your child's school to install monitoring software to think about this. A useful question to sit with is simpler: of the work my child is actually handing in, how much of it is a document that could, in principle, have been generated by someone who never engaged with the material at all? That's not an argument against essays entirely, writing clearly is still a real and valuable skill. It's a reason to make sure a written assignment isn't the only kind of proof your child is ever asked to produce.

If you'd like to see what assessment looks like when the deliverable is a real product instead of a document, something your child can run, explain, and defend because they actually built it, book a free consultation. We'll show you exactly how that works across our 24-week programme, from a student's first idea to something real they can stand behind.