How Teachers Detect AI in A-Level Film Studies Coursework

In A-Level Film Studies coursework, the risk of AI use sits mostly in one place. The coursework is not one thing - it is a creative production (a short film or screenplay) paired with a written evaluative analysis. A student cannot ask a chatbot to shoot their film for them, but they absolutely can ask one to write the reflective evaluation, the analysis of their creative decisions, or the theoretical framing that connects their work to the wider study of film. That written component is where generative tools like ChatGPT and Claude quietly appear, and it is where a fair, evidence-led detection routine matters.

The advice below is for UK teachers marking WJEC/Eduqas or other A-Level Film Studies specifications. Throughout, an AI likelihood score is treated as the beginning of a conversation, never the end of one.

Where AI Shows Up in Film Coursework

Film Studies evaluations reward a specific, personal voice: the student is meant to explain the choices they made, why a particular framing or edit serves their intention, and how their production sits against the auteurs and movements they have studied. Generative AI is good at producing fluent, confident-sounding film analysis in the abstract - but it does not know what your student actually filmed. That gap is the tell.

• Generic theory, no production. Polished paragraphs about mise-en-scène or Todorov's narrative theory that never connect to the specific shots the student made.
• Confident claims about scenes that do not exist. AI will happily invent a "low-angle tracking shot" the student never filmed.
• A voice that does not match their lessons. An evaluation written in flawless academic register from a student whose seminar contributions and drafts read very differently.
• Even coverage with no reflection. Real evaluations are lumpy - students dwell on the decisions they cared about. AI output is uniformly smooth.

None of these on their own proves anything. Together, and alongside a likelihood score, they build a picture you can act on responsibly.

What an AI Likelihood Score Actually Tells You

AI detection is probabilistic, not forensic. A tool reads the writing and returns a likelihood - an estimate, on a 0-100% scale, of how AI-generated the text appears. It does not produce proof, and it should never be treated as a confession. GradeOrbit's AI detection, part of its school accounts, is designed around exactly this framing: it gives you a score from 0 to 100 showing how far a piece of written work departs from that student's own earlier writing, with a confidence label and its reasoning, so you can decide where to look more closely - not a green-light/red-light verdict that decides a malpractice case for you. If you want the mechanics, our explainer on how AI detection likelihood scores work walks through what the number means and does not mean.

The professional discipline is to use the score to direct your attention. A high likelihood on a Film Studies evaluation is a prompt to re-read that evaluation against the student's actual production, their earlier drafts, and what you have seen of them in lessons - not a prompt to accuse.

Handling a High Score Fairly

When a score comes back high, slow down. The JCQ's guidance on the use of AI in assessments is clear that centres must be confident a candidate's work is their own, and that any concern is handled through the proper malpractice process with evidence - not on the strength of a single automated output. That means corroboration.

Cross-check the evaluation against the film the student actually submitted: do the analytical claims describe real shots and edits? Look at draft history and the conversations you had while they were producing. Talk to the student about their decisions - someone who genuinely made the film can explain why they framed a scene the way they did; a chatbot cannot brief them on that.

Keeping Detection Fast and Recording It

A whole class set of coursework evaluations is a lot of reading, and the point of a detection tool is to help you triage it rather than add hours. On a GradeOrbit school account, you can run a whole class's written work through detection at once - including scanned handwritten drafts, which the AI reads straight from the page images - so you can see at a glance which evaluations warrant a closer, corroborated read. Under your school's signed Data Processing Agreement (DPA), student work is held for your school's agreed retention window so detection can compare each evaluation against that pupil's earlier work on file; a pupil with no earlier work on file is skipped rather than given a generic score. Individual teacher accounts do not include detection; they cover marking, where uploaded work is redacted client-side and never stored.

Whatever you find, record your reasoning. A short note of the likelihood score, the specific mismatches you spotted between the evaluation and the production, and the outcome of your conversation with the student gives you defensible, professional evidence if a case is ever escalated to a moderator or examinations officer. Detection is one input into your judgment, and your judgment is what you document.

From Score to Decision

For a Film Studies class set, the routine is short: run the written evaluations through detection, use the likelihood score to decide which ones to re-read, check each of those against the film or screenplay the student actually made, talk to the student where the two do not match, and write down what you found. The score narrows your attention; the decision is yours.

Detection runs on GradeOrbit for Schools accounts and is included with each teacher seat.