How Teachers Detect AI in A-Level D&T Coursework

It is tempting to think of Design and Technology as a practical, hands-on subject that sits outside the reach of generative AI. The student designs a product, makes it, tests it. But A-Level D&T - whether AQA's Product Design, the Pearson Edexcel route, or OCR's specification - is, in assessment terms, a heavily written qualification. The non-exam assessment is a substantial design portfolio: a design brief and investigation, research into existing products and user needs, a specification, the development and modelling commentary, and a detailed evaluation against the brief. A large share of the marks sits in extended written analysis produced at home, over months, away from your supervision. That is exactly the kind of writing a model reproduces convincingly. Learning to detect AI in A-Level D&T coursework means knowing which parts of the portfolio are exposed, and what a genuine design process leaves behind that a model cannot invent.

This guide is for A-Level Design and Technology teachers marking the written elements of the design and make project.

Which Parts of D&T Coursework Are Exposed

The made outcome is the most protected component: you saw it built, you supervised the workshop time, you know whose hands were on the tools. The risk lies in the writing that surrounds the product. The contextual research, the analysis of existing products, the user-needs investigation, the justification of design decisions, and above all the final evaluation against the specification - these are written away from your eye and graded on the quality of reasoning and analysis. A model asked to "write a critical analysis of three existing products against a design brief for a sustainable desk lamp" will produce something fluent, structured, and full of the right design vocabulary: ergonomics, anthropometrics, sustainability, manufacturing feasibility.

This is why the written portfolio deserves the same scrutiny you would give an essay subject. Polished design prose is not evidence of a genuine investigation - and in D&T, where the language of analysis and evaluation is well rehearsed and widely published, a model has an enormous amount of material to imitate.

What a Real Design Process Leaves Behind

The single strongest check in D&T is whether the writing is anchored in the specific product this student actually made. A real design journey is messy and particular. A genuine portfolio references the decisions that actually happened: the prototype that warped, the material that was too expensive and had to be swapped, the user feedback that sent the design in an unexpected direction, the joint that failed testing and was redesigned. AI-generated analysis defaults to the generic - it describes the idealised, textbook version of a design process, not the awkward, contingent one that played out at your benches.

You have an advantage an English or History teacher does not: you were there. You watched the modelling, you saw the iterations, you know which student struggled with the CAD and which reworked the chassis three times. When a portfolio describes a smooth, theoretically perfect development that does not match the product you watched take shape, that gap is more telling than any score. Cross-checking the written commentary against the artefact you supervised is the most concrete check available, and one fluent prose cannot pass.

Look closely, too, at the evaluation. A student who genuinely tested their product against the specification writes about the specific failures and compromises: the dimension that came out wrong, the function that did not quite work, the user who found the handle awkward. A generated evaluation tends to be neatly balanced and oddly conclusive - praising the product against every criterion in turn without the honest admissions a real maker includes.

Reading the Likelihood Score

Your knowledge of the build tells you about the substance; an AI detection tool tells you about the prose. The two together are far stronger than either alone. On a school account, GradeOrbit's AI detection returns a score between 0 and 100 on the body of the writing - a probabilistic measure of how far the text departs from that student's own earlier writing on file. A high score on a contextual analysis, combined with research that does not connect to the product you watched being made, points clearly in one direction. A modest score on writing that captures the real, specific decisions of the build is reassurance, not suspicion.

As we explain in our guide on how AI detection likelihood scores work, the number is one input into a picture, never a verdict on its own. For D&T, the workshop eyewitness check is the second, very tangible input - and one you can ask the student about directly. AI detection is included in every school teacher seat, whether it's a routine portfolio section or the final evaluation heading into the assessed mark - there's no per-check charge and no tier to choose. It works by comparing the submission against that pupil's own earlier writing on file, which gives you a more useful signal than a generic score ever did. A pupil with no earlier work on file is skipped rather than given a generic score.

Turning Concern Into a Fair Conversation

When both the prose and the content raise concerns, the next step is a conversation, not an accusation. D&T gives you an unusually good opening because the work is physical, recent, and specific. "Talk me through why you changed the material here" is a simple, reasonable question. A student who lived the process can describe the false starts, the constraints, the reasons behind their choices. A student who generated the analysis will struggle to connect the writing to anything that happened at the bench.

Keep it curious rather than confrontational. Ask about the prototype that failed, the user feedback they claim to have gathered, the manufacturing process they specified. The point is to move from your impression of the writing to factual questions about a design process that genuinely took place. The broader framework sits in how teachers handle a high AI likelihood score fairly. Detection should never replace your professional judgment; a high score on an analytical genre is a prompt to look at the work, not a conviction.

The EAL and Lower-Confidence-Writer Caveat

Technical analysis rewards a measured, slightly impersonal register - exactly the style detection tools can over-associate with AI. A careful student who has absorbed the conventions of design writing, or an EAL student writing in a deliberately neutral voice, may produce prose that scores higher than their actual authorship warrants. This is why the workshop check matters so much in D&T: it gives you evidence independent of writing style. A student whose portfolio matches the product you watched, and who can talk through their design decisions, has done the work, whatever the prose score suggests. Never let a likelihood score alone override that.

Documenting the Check for Moderation

Whatever you conclude, record it at the time. For a D&T portfolio this is straightforward: note which written section you checked, whether the commentary matched the product you supervised, the likelihood score on the prose, what you asked, and what the student said. "Contextual analysis likelihood 84%; describes user research not evident in the development folder; student unable to explain the material change when asked" is a clear, factual record that holds up if the case is revisited. For coursework feeding into internally assessed and moderated NEA components, keep the detection run and your workshop notes alongside your marking, so the evidence pack reflects how you reached the mark. If your school's policy does not yet address AI specifically, the school AI academic integrity policy guide is a useful starting point.

Running D&T Coursework Through GradeOrbit

AI detection runs on school accounts, where pupils' work is held under the school's Data Processing Agreement for its chosen retention window and compared with their own earlier writing. Run it from a pupil's profile, or on the whole group at once. For typed portfolios, export to PDF (Word documents aren't accepted); for handwritten design folders and annotated sheets, scan or photograph the pages clearly, or use the phone camera link. For a class set, bulk upload the pages with the number of pages per student, and GradeOrbit reads the name on each cover page and matches it to your class list - you pick the pupil if the match isn't clear. Results appear on each pupil's profile with the score, a confidence label, the signals noticed and a reasoning paragraph. The record of the detection, your workshop notes, and your decision belongs in your own records and your school's systems. For detecting AI in the GCSE end of the subject, our guide on detecting AI in GCSE Design and Technology coursework applies the same principles to younger students.

Try GradeOrbit for AI Detection in D&T Coursework

If you want a reliable AI detection tool that gives you a clear likelihood score to work alongside your own knowledge of the build and your professional judgment, GradeOrbit school accounts are built for exactly that, with detection included in every teacher seat.

See how detection works on a school account and start a free school trial to try it on real portfolio writing before any commitment.