Should Schools Use AI Detectors for GCSE Coursework?

Detection software promises certainty about whether a pupil used AI. It cannot deliver it — and using it as though it can risks serious unfairness to pupils. Here's the honest position, and what works better.

The short answer

No — not as the basis for any decision about a pupil. AI detectors are not accurate enough to support an academic misconduct allegation, and using them that way creates real risk of serious injustice to pupils. They can have a very limited role as one weak signal among many, but a school that treats a detector score as evidence is on dangerous ground.

Why detectors are unreliable

They produce false positives

Genuine pupil work is regularly flagged as AI-generated. Since a false positive means accusing an innocent pupil of cheating — with consequences for their record, their grades and their trust in school — even a low false positive rate is unacceptable for high-stakes use.

They are biased against some pupils

This is the most serious problem. Detectors disproportionately flag work by pupils writing in a second language, and by pupils whose writing is simple and formulaic. Those are often precisely the pupils least able to defend themselves against an accusation.

The scores are not what they appear

"87% AI" is not a probability that a pupil cheated. It's an opaque model's confidence score, presented with a precision the underlying method doesn't justify.

They are easily defeated

Light editing, paraphrasing tools or simply asking AI to write less formally will defeat most detectors — so the pupils determined to cheat are the ones least likely to be caught, while the honest pupil with an unusual style gets flagged.

What about coursework and NEA?

For work contributing to a qualification, follow JCQ's guidance on AI in assessments and your exam board's requirements — those set out the authentication process expected, which is built around teacher declaration and supervision rather than detection software. Where possible, supervise more of the work in class and keep drafts; that authenticates far more robustly than any tool.

What to do instead

• Know your pupils' work — a mismatch with prior writing is the strongest signal there is
• Assess process — plans, drafts, and evidence of the journey
• Talk to the pupil — ask them to explain their reasoning; authors can, others can't
• Supervise key writing — in-class conditions for anything that really matters
• Design tasks where AI doesn't bypass the learning
• Set clear expectations so pupils know what's permitted and what must be declared

See the signals that genuinely indicate AI use.

If your school does use a detector

Then be clear internally that it is a prompt to look more closely, never a verdict. Write that into your policy, make sure every member of staff understands it, and never quote a score to a pupil or parent as though it were evidence. See writing your AI policy.

The bigger point

Detection is an arms race schools will not win, and pursuing it damages the relationship of trust that makes a school work. The schools handling this best have stopped trying to catch pupils and started redesigning assessment so that honest work is the path of least resistance.