What breaks when students write essays with AI?
You teach or lead a program, assign an essay, and get back fluent machine prose. Now you cannot tell who learned what, and accusing the wrong student costs more than missing a cheater.
Coursework stopped measuring learning
The Office for Students says AI has made it close to impossible to assess university coursework, and the credibility of English degrees is under threat. The Guardian reports the OfS wants institutions to return to traditional exams and in-person presentations. The Observer notes the share of British students using AI for assignments has risen sharply, and the regulator treats this as permanent. The essay, once the core proof of thinking, now proves only access to a chatbot.
- The OfS urges written exams and vivas instead of coursework.
- The Observer reports the trend threatens to devalue degrees.
Detection tools punish the innocent
US campuses are shunning AI detection software altogether, The Daily Star reports. Per Digital Trends, a Bloomberg investigation found tools like GPTZero and Copyleaks wrongly flagged pre-ChatGPT essays as AI-written, and one UC Davis senior received a zero on a midterm he wrote himself. The false-positive cost is now seen as worse than the cheating detectors might catch.
Redesign assessment, don't police it
The University of Western Australia rejected detection tools because they would punish the majority of honest students, the AFR reports. Its chancellor, Diane Smith-Gander, said cheating existed long before AI, so UWA is rethinking assessment to protect genuine learning. Some writing programs already do this, Digital Trends reports, by collecting handwritten outlines and video reflections instead of one polished essay. The goal is not catching cheaters, it is designing tasks AI cannot fake.
What this does not solve
The Guardian and Observer both report institutions retreating to written exams and vivas. Those are costly, stressful, and the supplied sources do not claim they restore creative learning. University World News argues AI policies must also address data, not only cheating. In-person assessment may protect degree credibility without fixing why students outsource thinking in the first place.
Which of your current assessments could a chatbot complete without the student thinking?
Could you prove each student's learning with evidence produced in the room?
Pick one assignment you grade and ask: could AI produce this answer in ten seconds?
Sources
Recommended for you
- Education
Why does plagiarism software still miss actual cheating?
Detection tools compare text against other text. That design choice means they catch copying, not the dishonesty you actually care about.
- Education
What breaks when AI tutors outperform your teachers?
A Harvard-MIT trial shows AI tutoring beats best-practice classrooms. The risk is not the tool, it is who designs how it gets used.
- Education
What breaks when universities grade AI-era essays?
AI writes passing essays, so written assessment no longer proves learning. Universities are falling back on oral exams, but scaling them is the real problem.
- Education
Why does AI widen learning gaps before closing them?
Students use AI everywhere, but almost none are taught how. The gap that matters is guidance, not access, and it splits learners fast.
Enjoyed this article?
Subscribe for new articles. No spam. Unsubscribe anytime.