Spotting AI in Year 10 End-of-Year Coursework Drafts
End of Year 10 is one of the most overlooked points in the GCSE timeline for AI misuse. Students have just enough coursework experience to know what a "good" draft looks like, exam pressure has not yet kicked in, and the summer holidays sit between them and the final Year 11 submission window. A draft that quietly leans on ChatGPT or Claude in May or June can resurface in September as the student's working version - and by then the voice has been laundered through six months of redrafting.
Spotting AI in Year 10 end-of-year coursework drafts is therefore less about catching cheating and more about catching drift early, while there is still time to coach the student back to their own voice before high-stakes submission.
Why End-of-Year Drafts Are an AI Hotspot
Year 10 coursework drafts share a few features that make them especially vulnerable to AI use. The grade does not yet count toward the final qualification, so students underestimate the seriousness of using a chatbot. Teachers are deep in Year 11 mock marking and end-of-year reports, so draft feedback turnaround slows down. Parents are nudging children to "get ahead over summer", and the gap between "use AI to brainstorm" and "use AI to write" is much narrower than most 15-year-olds realise.
The result is a draft that arrives polished beyond the student's classwork: tightly argued, evenly paragraphed, with vocabulary that the student has never used out loud. Without a structured way to check, these drafts feed into Year 11 with their provenance unexamined.
What an AI Likelihood Score Actually Means
AI detection is probabilistic, not forensic. Most generic detectors return a likelihood score between 0 and 100 percent - an estimate of how AI-like the text patterns are, based on a model trained on a large corpus of human and AI writing. It is not a verdict and it does not replace professional judgment.
On a generic detector, a 92 percent score does not mean "92 percent of this was written by AI". It means "the patterns in this text are highly characteristic of AI-generated prose". A 38 percent score does not mean "38 percent AI" - it means the patterns are mixed, and the most likely interpretation is heavy AI editing of a human draft, or a human writer mimicking AI-style structure they read online.
Treat the score as a triage tool. It tells you where to look more carefully. It does not tell you what to do.
Handling Borderline Scores Fairly
Anything between roughly 30 and 70 percent is borderline territory. A useful next step is to look at the student's classwork from the same term and compare on three axes: vocabulary range, sentence rhythm, and idea density. Year 10 students grow as writers, but the growth is gradual. A sudden shift in all three at once is the strongest signal that something external has shaped the draft.
Read the borderline draft aloud. AI-generated prose tends to be confident and evenly weighted across every paragraph. Real Year 10 writing has uneven energy - strong openings that thin out, an idea that gets repeated because the student forgot they had already made it, a closing paragraph that visibly tires. If the draft reads with suspiciously even stamina, that is worth a conversation.
Acting on High Scores Before Year 11
Where a Year 10 draft scores 80 percent or higher and the comparison with classwork supports the signal, the conversation needs to happen now - not in September. The framing matters. This is not a misconduct meeting in May; it is a coaching conversation about what is and is not allowed when the work starts counting in Year 11.
Ask the student to talk you through their planning notes. Ask them to explain a specific paragraph in their own words. Where the draft was largely AI-generated, the student usually cannot do this - and that gap itself is the teaching moment. Help them rewrite a section from scratch with you in the room, so they have a baseline of their own voice to anchor the rest of the year to.
Document the conversation briefly in your records. A short note - date, score, comparison with classwork, action agreed - is enough to support any later conversation with the head of department or with the student's Year 11 teacher.
Where GradeOrbit Fits
GradeOrbit's AI detection is built for exactly this kind of low-drama, mid-term triage. It runs on school accounts, unlimited within every teacher seat: upload Year 10 drafts as scans, photos or PDFs, run the check from a pupil's profile or on the whole class, and GradeOrbit compares each draft against that pupil's own earlier work already marked in GradeOrbit. The result is a score from 0 to 100 for how far the draft departs from their own writing, with a confidence label, the signals it noticed, and a plain-English reasoning paragraph - a stronger signal than a generic score for spotting drift early. A pupil with no earlier work on file is skipped, and you are told.
Pupils' work is held under the school's Data Processing Agreement and never used to train AI models. Because it sits under the school's agreement, with a set retention window and deletion at offboarding, there is no unmanaged record of student work sitting on a third-party server through the summer holidays - a detail that matters for school data protection officers as much as for teachers.
Try GradeOrbit for Your End-of-Year Drafts
If you are heading into the end-of-term marking pile and want a fast, defensible way to flag AI use in Year 10 coursework drafts before they harden into Year 11 submissions, GradeOrbit's school accounts are built for it. Start a free school trial and start screening drafts.