Detecting AI in GCSE Hospitality & Catering Coursework

Hospitality and Catering is one of the more exposed vocational assessments on the timetable. The written components - dish planning, nutritional analysis, provenance research, and evaluative reflection on practical outcomes - are exactly the sort of structured, criterion-led writing that ChatGPT and Claude produce quickly and fluently. A student can generate a plausible-looking rationale for a menu in seconds, and the marking grid rewards precisely the qualities a language model is good at faking: clear structure, sensible terminology, and coverage of every assessment point.

For the teacher marking that work, three things matter: what a probabilistic likelihood score actually measures, what to do when a piece of coursework scores high, and how to keep your own professional judgment as the assessor at the centre of every decision.

Why Catering Coursework Is Easy to Draft With AI

The WJEC/Eduqas Level 1/2 award and the various GCSE-equivalent catering qualifications share a common shape: students write about food before, during, and after they cook it. They justify choices, analyse nutrition, cost a dish, and reflect on what went wrong and how they would improve. Each of those tasks has a predictable structure, and predictable structure is what AI writing does best.

The result is that a moderation pile can contain submissions that read fluently and tick every box, yet come from a student who struggles to explain a single sentence when you ask them at the hob. The writing and the practical skill have come apart - and the written portfolio no longer reliably tells you what the student actually knows.

The Evaluation Section Is the Tell

Evaluative reflection is where AI writing shows most. Genuine evaluation is specific: it names the dish that split, the sauce that was too thin, the timing that slipped because the oven was full. AI-drafted evaluation tends to be generic - "the dish could be improved with better time management" - because the model was never in the kitchen. When you read a reflection that could apply to any student's practical session, that is worth a closer look.

How Likelihood Scores Actually Work

AI detection does not return a yes-or-no verdict, and you should be suspicious of any tool that claims to. GradeOrbit's detection tool, available on school accounts, returns a likelihood score from 0 to 100 - a measure of how far the new work departs from that student's own earlier writing on file, with a Low, Medium or High confidence label, the signals it noticed and a plain-English reasoning paragraph.

That probabilistic framing matters enormously in a fairness-sensitive subject. A score of 82 does not prove a student cheated. It tells you the sentence rhythm, vocabulary spread, and predictability of word choice differ from this student's earlier writing in ways worth a closer look. You then bring what you know - the student's practical ability, their previous portfolio work, the assignment brief - to interpret what that number means.

What a Detection Check Costs

On a GradeOrbit school account, unlimited AI detection is included in every teacher seat, whether it is a routine pass across a whole class portfolio or a single piece you are escalating for closer scrutiny before a malpractice conversation, with no per-check charge. The check itself is baseline-aware, comparing the submission against that student's own earlier work on file under your school's Data Processing Agreement (DPA), which is a stronger signal than a generic score regardless of how much scrutiny the piece needs; a student with no earlier work on file is skipped, and you are told. You can run it from a single student's profile or across the whole class. A sensible pattern for catering is still to run every submission at the coursework deadline, then reserve your own closer read for the handful that flag highest.

What to Do When a Portfolio Scores High

A high score is the start of a conversation, not the end of one. Before anyone completes a malpractice form, do three things.

First, compare against the student's own track record. Catering teachers usually have months of evidence - earlier written tasks, practical observation notes, planning sheets. Writing that is wildly out of line with everything else a student has produced is far more concerning than a high score on work that matches their usual standard.

Second, talk to the student. Ask them to explain a specific paragraph in their own words - why they chose that cut of meat, why that cooking method, what the nutritional trade-off was. Genuine writers can always explain their choices; students who leaned on AI usually cannot, and that conversation recorded in your notes is often stronger evidence than the score itself.

Third, apply your professional judgment. You are the assessor. The score is one data point among several.

Building a Consistent Department Approach

If you lead catering delivery, agree a shared threshold with your team before the next deadline - a score of 70 as the trigger for a verifier conversation, say, or a stricter 85. The exact figure matters less than the consistency - every student is treated the same way, and every flagged piece goes through the same evidence-gathering process. Tell students at the start of the unit that detection is part of how submitted work is checked; that transparency reduces AI use far more effectively than punishing it after the fact.

How GradeOrbit Handles Student Data

Because AI detection runs on a school account covered by a signed Data Processing Agreement (DPA), students' work is held under that agreement with their real names, so each check can compare against that student's earlier work on file. Uploaded scans are removed when a run archives (after 90 days by default, and the school can set anything from 7 to 365 days), student records follow your school's retention window, and everything is wiped at offboarding. Individual teacher accounts do not include detection; they cover marking, and GradeOrbit does not store the student work you upload there: work is redacted client-side in your browser, students are referred to only as "Student 1", "Student 2", and so on, and uploaded work is discarded immediately after marking. Across all tiers, student work is never used to train AI models. GradeOrbit is an assistive tool: it gives you evidence and a starting point; it never replaces your judgment as the assessor, and it never makes the malpractice decision for you.

In Short

Keep your practical observation notes and planning sheets to hand, because the evaluation section is usually where the clearest evidence sits: a reflection that could describe anyone's practical is the one to ask about. A detection score helps you decide which portfolios to read closely; the conversation at the hob and your knowledge of the student decide the rest.

Details of the school account and how to start a free school trial are on the GradeOrbit for Schools page.