How Teachers Detect AI in A-Level History Coursework

The A-Level History coursework essay - the non-examined assessment, or NEA, that runs to three to four thousand words across the AQA, Edexcel, and OCR specifications - is one of the highest AI-risk pieces of work a student will submit during their two years of study. It is written largely at home, over weeks, with access to every research tool available, and it asks for exactly the kind of structured historical argument that AI tools produce fluently. If you teach the subject, checking NEA essays for AI use is now part of the job, and the process has to be fair and defensible - which means never treating a detection score as a verdict.

Why the History NEA Is Especially Exposed

The coursework component of A-Level History is built around independent enquiry: students choose or are assigned a question, read across a range of primary and secondary sources, and construct a sustained analytical argument that reaches a substantiated judgement. The conditions that make it a rich piece of academic work are precisely the conditions that make AI misuse plausible. There is no exam hall, no time limit, and weeks of working time during which a draft can be generated, edited, and made to look like the product of genuine research.

The subject content compounds the risk. The historiography of widely studied topics - the causes of the First World War, the nature of Tudor government, the civil rights movement, Stalin's Russia - is extensively documented online, which means an AI tool can produce a fluent, well-organised, superficially convincing essay on most NEA questions with very little prompting. That fluency is exactly what makes AI-generated history difficult to spot on a first read, and exactly why a detection tool gives you useful evidence where your own instinct might not immediately flag anything.

None of this means most students are misusing AI. The overwhelming majority are not. But the coursework's high stakes - it contributes a significant proportion of the final A-Level grade and feeds into university applications - mean that running detection as a routine, fair check protects the integrity of the qualification for the students who did the work honestly.

What GradeOrbit's Likelihood Score Tells You

GradeOrbit's AI detection, on school accounts, returns a score from 0 to 100 on every piece of work you check, with a Low, Medium or High confidence label, the signals it noticed and a plain-English reasoning paragraph. The score measures how far the new work departs from that student's own earlier writing on file: low means consistent with it, high means it differs in ways worth a closer look. It does not tell you with certainty that a particular student used AI. It tells you where the balance of evidence sits, and how unusual the writing is relative to what this student has produced before.

For the History NEA there are two things worth holding in mind when you read a score. First, a capable A-Level historian who has genuinely absorbed the historiography and writes in a polished, academic register will naturally share some surface features with AI output - strong structure, formal vocabulary, confident argument. A high score on such a piece is not proof of misuse. Second, the coursework is long, which gives the model a substantial body of text to work with and generally makes the score more informative than it would be on a short paragraph. Both factors point the same way: the score is a prompt to look more closely, not a conclusion to act on.

On a GradeOrbit school account, unlimited AI detection is included in every teacher seat, whether you're triaging a set of drafts or checking a final NEA submission heading into moderation, with no per-check charge. What makes the check reliable for coursework of this weight is that it's baseline-aware: under your school's signed Data Processing Agreement (DPA), it compares the submission against that student's own earlier writing on file, rather than scoring it against a generic "sounds like AI" measure.

Reading a High Score in Context

A score above 70% on an A-Level History coursework essay warrants a closer look. The most useful comparison you have is the student's own work produced under your supervision: timed essays from class, exam practice, source-analysis tasks completed in lessons. Read the flagged coursework alongside that supervised writing. Does the argument sound like the same student? Is the analytical sophistication consistent, or does the coursework leap to a level of fluency and historiographical command that their timed work never approaches?

Genuine A-Level history writing tends to carry the fingerprints of the way it was taught. Students lean on the interpretations you emphasised, cite the historians you put in front of them, and reproduce the particular debates your lessons foregrounded. They also make characteristic errors - misremembering a date, overstating a historian's position, leaning too hard on a single source. AI-generated history is often more evenly competent and oddly unanchored: technically accurate, smoothly argued, but detached from the specific texts and interpretations your class actually worked with. A coursework essay that cites historians you never taught, in a register the student has never previously produced, is worth investigating when the score is also high.

If the flagged piece is broadly consistent with the student's supervised work and the score is merely elevated, the number is most likely reflecting a strong, genuine historian rather than AI generation. The score is one input - the rest of the picture comes from what you know about the student and the work.

Starting a Fair Conversation With the Student

If both the score and your own read raise concerns, the next step is a conversation, not an accusation. The most reliable approach is to open with the history, not the detection result. Ask the student to talk you through a specific part of their argument: "Walk me through why you weighted this interpretation over the other one here," or "Which source led you to this judgement, and what made it convincing?" A student who genuinely researched and wrote the essay can discuss the reasoning behind it. A student who generated it usually cannot reconstruct the argument in their own words, or gives an account that does not match the sophistication of what they submitted.

Keep the tone low-key and specific. You are not running a tribunal; you are doing what any history teacher does when a piece of work surprises them - asking the student to demonstrate the thinking behind it. Students who did the work talk about it with the confidence and the gaps you would expect from someone who has wrestled with a difficult question over several weeks. That texture is hard to fake and very revealing.

Documenting Your Decision for Moderation

A-Level History coursework is internally marked and externally moderated, which makes contemporaneous documentation especially important. Whatever your conversation concludes, write it down at the time. A few sentences will do: the score, the nature of your concern, what you asked the student, what they said, and the decision you reached. A record made on the day is far more defensible than one reconstructed from memory months later if the case is revisited.

If you are satisfied the work is the student's own, note that explicitly - "Likelihood score 74%; consistent with supervised essays; student explained the historiographical choices fluently in conversation; no further action." If concerns remain, follow your centre's academic integrity policy and the awarding body's malpractice procedures for the next step. Including the detection run in your marking notes shows a moderator that due diligence was applied; the moderation cycle guide covers how detection results sit within the wider evidence pack for internally assessed components.

Running Coursework Through GradeOrbit

For an NEA essay, run the check from the pupil's profile on a school account, or upload a class set to the class: bulk upload splits the pages by pages per student, and GradeOrbit matches the name on each cover page to the class list, asking you to pick the pupil where the match isn't clear. The essay is then read against the pupil's own earlier work on file, which is the comparison that matters most for a long piece of independent work; a pupil with no earlier work on file is skipped, and you are told. For typed coursework, export the document to PDF and upload it (Word files aren't accepted); for handwritten drafts, scan or photograph the pages clearly and upload the images, or use the phone QR camera link. The AI reads the page images directly, so the format of the work does not matter. The same workflow applies whether you are checking a single final submission or triaging a class set, and it mirrors the approach in our guide to detecting AI in GCSE History coursework.

Because AI detection compares an NEA essay against earlier work on file, it runs on GradeOrbit school accounts under a signed Data Processing Agreement (DPA), where pupils' work is held with their real names: uploaded scans are removed when a run archives (after 90 days by default, adjustable from 7 to 365 days), pupil records follow your school's retention window, and everything is wiped at offboarding. Individual teacher accounts do not include detection; they cover marking, where uploaded student work is redacted client-side in your browser, labelled anonymously, and never stored (processed and discarded immediately). Across every tier, student work is never used to train AI models.

Testing Detection on Essays You Already Know

The fairest test of any detection tool is a set of NEA drafts from students whose supervised writing you know well: you can see whether the scores line up with what you would expect before relying on them for a decision. GradeOrbit for Schools includes AI detection in every teacher seat, and a free school trial gives you time to run one History set through it.

Seats, the Data Processing Agreement and retention settings are explained at gradeorbit.co.uk/schools.