How Teachers Detect AI in GCSE Engineering Coursework

GCSE Engineering is a practical subject on paper, but a surprising amount of the assessed work is written - and that is where a chatbot can slip in unnoticed. The design-and-make task asks students to justify decisions, analyse a brief, explain material choices and evaluate a prototype in extended prose. The maths, the sketches and the CAD are hard to fake; the paragraphs of rationale and evaluation around them are not. That written scaffolding is where AI-assisted work tends to appear.

Where the Written Marks Actually Sit

Across the main GCSE Engineering specifications, marks are spread over investigating a brief, developing and communicating design ideas, planning manufacture, and analysing and evaluating outcomes. Several of those strands are assessed through writing: the analysis of the problem, the justification of material and process choices, the reasoning behind a design decision, and the reflective evaluation of the finished product against the specification. A student can build a perfectly good prototype and then hand in written sections that read nothing like their workshop conversations - polished, generic and oddly confident.

That mismatch is the thing to notice. AI is fluent at the surface features of good engineering writing - it will happily explain why aluminium was chosen over mild steel, or describe a risk assessment in textbook terms - but it does not know what happened at the bench. Genuine coursework carries the fingerprints of the actual project: the tolerance that was missed, the jig that had to be remade, the design that changed after the first test.

What AI-Assisted Engineering Writing Tends to Look Like

There are patterns worth reading for, as long as you hold them lightly and look for several at once rather than seizing on any single one.

Textbook Justification, No Project

AI-written rationale tends to explain material and process choices in general terms that would fit any brief - "aluminium offers a good strength-to-weight ratio and resists corrosion" - without tying the decision to this student's specific product, dimensions or user. Real justification is anchored to the task: it references the actual load, the actual budget, the actual constraint the student was working within.

Evaluation With Nothing to Evaluate

The evaluation section is a giveaway. A chatbot can produce a balanced-sounding reflection on strengths and improvements, but it has no memory of the build, so it stays abstract. Authentic evaluation names what actually went wrong, what the testing showed, and what the student would change with hindsight - messy, specific and often self-critical in a way generic prose is not.

Vocabulary Above the Bench

Watch for technical language that outruns the rest of the folder - a sudden fluency with manufacturing terminology or standards that never appears in the student's sketch annotations, planning notes or lesson contributions. A step-change in register between the practical evidence and the written analysis is worth a second read.

How AI Detection Likelihood Scores Work

It helps to be clear about what a detection tool actually gives you. GradeOrbit's AI detection, on school accounts, does not return a yes-or-no verdict; it returns a score from 0 to 100 that measures how far the new writing departs from that student's own earlier work on file, with a confidence label, the signals it noticed and a plain-English reasoning paragraph. That is an indicator, not proof. A high score means the writing differs from the student's usual work in ways worth a closer look; it does not, on its own, establish that a student cheated.

This matters most for the students who write in a naturally formal, well-organised way - often your most diligent - because clean, structured prose can read as machine-like to a detector. Treated as a single input alongside your professional knowledge of the student, the folder and the workshop, a likelihood score is genuinely useful. Treated as a verdict, it will eventually be unfair to someone who did nothing wrong. Our guide on how AI detection likelihood scores work goes deeper on reading the number well.

Handling a High Score Fairly

A high likelihood score is the start of a conversation, not the end of one. The fair next step is to gather corroborating evidence rather than to accuse. Compare the flagged writing against the student's controlled or in-class work, look at whether the design evidence and the written rationale tell the same story, and check version history or draft stages where they exist. Then talk to the student about their process - asking them to walk you through a design decision usually tells you far more than the score did.

On a GradeOrbit school account, unlimited AI detection is included in every teacher seat, regardless of document length, with no per-check charge. A sensible routine is still to run it across a class set and reserve your own closer read for the pieces that genuinely stand out, because under your school's signed Data Processing Agreement (DPA), the check compares the writing against that student's own earlier work on file rather than returning a generic one-off score; a student with no earlier work on file is skipped, and you are told. Individual teacher accounts do not include detection; they cover marking, where uploaded student work is redacted client-side in your browser and never stored. For the wider process of documenting and acting on scores, our piece on how teachers handle a high AI likelihood score fairly sets out a defensible approach.

Keeping Your Judgment at the Centre

Detection technology is assistive, not authoritative. It exists to draw your attention to writing worth a second read, not to make the decision for you. For a subject like Engineering, where the whole point of coursework is that a student designed, made and reflected on something real, your knowledge of that project is the strongest evidence you have. The likelihood score helps you decide where to look harder.

In practice that means uploading the written sections of a student's folder, reading the baseline-aware 0-100 score alongside what you know of the build, and keeping your closest reading for the pieces that stand out. You can start a GradeOrbit school account with a free school trial if you want to test this on a real set of GCSE Engineering coursework.