NEA or EPQ Flagged as AI Before You Submit? What the JCQ Rules Say

If the concern comes up before you sign the declaration of authentication, your school deals with it. JCQ's guidance says a suspected AI problem at that stage "does not need reporting to the awarding organisation and must be dealt with in the centre directly". Once you have signed, the school has to report it to the exam board. In neither case is a detector score the decision. What decides whether your work goes in is whether your teacher can sign to confirm it is your own.

JCQ, the Joint Council for Qualifications, is the membership body for the UK's main qualification providers; AQA, OCR, Pearson, WJEC and CCEA are among the members shown on its About page. Its rules apply to NEA, coursework and the EPQ. The main document here is "AI Use in Assessments: Your role in protecting the integrity of qualifications", dated 30 April 2025 and still the current version on 2 October 2026. JCQ's instructions to schools for NEA and for coursework, and its "Information for candidates" sheets, were updated for this school year and are effective from 1 September 2026.

Coursework under JCQ rules gets signed twice. You sign a declaration that the work is your own, and your teacher signs one too. The AI guidance tells schools to make sure teachers "are clear about their responsibility to only authenticate and submit work for assessment by the awarding organisation that they are confident is the student’s own".

JCQ's instructions for both NEA and coursework say what happens when a teacher cannot sign. The NEA version tells teachers to "record a mark of ‘0’ (zero) for internally assessed work"; the coursework version says "the candidate’s work cannot be accepted for assessment. A mark of ‘0’ (zero) must be submitted". That is the worst outcome at this stage, and it can be appealed. Schools must have an internal appeals procedure, and the coursework instructions say it "must cover appeals against decisions to reject a candidate’s coursework on the grounds of malpractice or not being able to confidently authenticate the candidate’s work."

The guidance places that judgement with the people who know you: "Teachers will know their students best and so are best placed to assess the authenticity of work submitted to them for assessment". The same sentence goes on, "AI detection tools can be a useful part of the evidence they can consider." So a score can count. It is not supposed to count on its own.

The guidance gives four examples: Copyleaks, GPTZero, Sapling and Turnitin AI writing detection, adding that "The list provided is not exhaustive". It says these tools "may be used as a check on student work and/or to verify concerns about the authenticity of student work", then adds two warnings of its own.

The first is that the tools "will give lower scores for AI-generated content which has been subsequently amended by students, as they base their scores on the predictability of words." The second is about accuracy, which it says varies "depending on the AI tool and version used, the proportion of AI to human content, prompt types and other factors (such as an individual’s English language competency)." If English is not your first language, note that JCQ lists your English level among the things that affect a detector's accuracy. We go through the research on that in why non-native English writers get flagged by AI detectors more often.

Detection "should form part of a holistic approach", and "all available information must be considered when reviewing any malpractice concerns."

There is no score threshold anywhere in the guidance. We searched the full text of the guidance page: "threshold", "percent" and "per cent" appear zero times, and the only percentage sign belongs to a case study in which a school estimated that 98% of a student's sketchbook content was influenced by AI. So JCQ sets no number that counts as a fail, and none that counts as safe.

The first row has two catches. JCQ's coursework and NEA instructions keep an exception for breaches of the assessment conditions, "e.g. possession of notes, communication with other candidates", which must be reported even before you sign. And the NEA instructions add that even when a concern is settled in school, "Details of any work which is not the candidate’s own must be recorded on the record form."

Every school also has to have a process for this. The guidance requires one "for teaching staff to follow where misuse of AI is suspected before the student has signed the declaration form", and JCQ's general regulations require a malpractice policy that covers AI. You can ask to read it.

After signing, the stakes change. The guidance says sanctions for plagiarism and a false declaration "range from a warning regarding future conduct to disqualification and the student being barred from entering for one or more examinations for a set period of time."

The guidance suggests comparing a piece of work with what you have written in class or under supervision, looking at "Spelling and punctuation", "Grammatical usage", "Writing style and tone", "Vocabulary", "Complexity and coherency" and "General understanding and working level".

Among its prevention steps it suggests that teachers "Examine intermediate stages in the production of work" and consider "a short verbal discussion about their work to ascertain they understand it and it reflects their own independent work".

It also lists 16 possible signs of AI misuse. They include American spelling used by default, references that cannot be found, warnings left over from an AI tool, and "Overly verbose or hyperbolic language that may not be in keeping with the candidate’s usual style." Some only make sense against your other work, such as "A difference in the language style used when compared to that used by a student in the classroom or in other previously submitted work." Your classwork and drafts are what that comparison is made with.

1. Ask what the concern rests on. Which tool, which sections, and whether a teacher has noticed something in the writing or there is only a number. You will answer it better if you know what it is.
2. Gather the record of how the work was made. Plans, notes, the sources you read, earlier drafts, feedback you were given, and for the EPQ your production log. Get version history out of Word or Google Docs as files you can show. We cover where each tool keeps it in how to keep version history in Word, Google Docs and Overleaf. Do not create a history after the fact; a document retyped today shows today's date.
3. Be ready to talk it through. JCQ names a short discussion as one way for teachers to check. Be able to say why you picked your sources, what you changed your mind about, and which part was hardest to write.
4. If you used any AI tool, check whether that use was allowed for this piece of work, and deal with it before you sign. JCQ's rule is that work "must be in their own words and not copied or paraphrased from another source such as an AI tool". Using AI to improve your own writing counts: in the first case JCQ publishes, an A Level History student who "admitted using ChatGPT for some of the content of their work, for both the improvement of their own work as well as the creation of entirely new content" lost all the marks for the NEA component. Translation now has its own rule. JCQ's 2026-27 instructions for both NEA and coursework add: "AI translation tools must not be used to generate work for assessment." In another published case, an A Level Art and Design student who used DeepL to translate source material into their sketchbook, knowing it was not allowed, got zero marks for the component. If any part of your work was translated with an AI tool, tell your teacher before you sign rather than relying on a reference. For AI use that is allowed, the student sheets for coursework and NEA ask for the same things: name the tool and the date, keep your prompts and the output in a non-editable form such as screenshots, and add a brief explanation of how you used it. "This must be submitted with your work for final assessment".
5. Ask before you rewrite, and keep every earlier version. For the EPQ, JCQ's coursework instructions say you "are free to revise and re-draft a piece of coursework without teacher involvement before submitting the final piece". For GCSE and A-level NEA the rules are tighter: unless your specification forbids it, teachers may, "having provided advice at a general level, allow candidates to revise and re-draft work", but "Teachers must not provisionally assess work and then allow the candidate to revise it." Either way, the earlier versions are part of your record.
6. Read your school's malpractice policy. It has to say how AI may be used, how to acknowledge it, and what happens when misuse is suspected before you sign.

The JCQ guidance applies whichever board sets your subject, but some of the boards' own documents are more specific about what to hand in.

Pearson's template predates the current JCQ guidance and refers to it by an older title, so ask your teacher whether your school uses it or its own log.

Can my school report me to the exam board over an AI score before I hand the work in? Under JCQ's guidance a concern raised before you sign the declaration does not need to be reported and is dealt with by the school. The exception is a breach of the assessment conditions, such as having notes you were not allowed.

Is this a new rule for 2026? Not for the AI guidance, which is still dated 30 April 2025. JCQ's 2026-27 instructions for coursework and NEA, effective from 1 September 2026, do add one new line: "AI translation tools must not be used to generate work for assessment." The 2026 student sheet repeats the requirement to keep your prompts and the AI output and hand them in with the work.

Does the EPQ come under these rules? Yes. JCQ's coursework instructions say coursework "includes the work required in Project qualifications". Those instructions do not cover GCSE and A-level NEA in England and Wales, which have their own NEA instructions.

If I acknowledge AI use, do I lose marks? For anything that relies on it, yes. The guidance says students who use AI "have not independently met the marking criteria therefore they will not be rewarded". In its A Level History example, work judged worth 22 to 24 marks got 20 because one section relied solely on an acknowledged AI tool. Losing marks for an acknowledged section is not the same as malpractice.

Can a detector score on its own get me disqualified? None of the six cases JCQ publishes rests on a score alone. Each also involved something else: a teacher, moderator or examiner noticing something in the work, a closer review against other indicators, notes compared with the final piece, or the student's own account. In the Pearson EPQ case the student "declined to provide a statement explaining the concerns" and was disqualified, so do not stay silent if you are asked.

• JCQ, AI Use in Assessments: Your role in protecting the integrity of qualifications (30 April 2025), read 2 October 2026.
• JCQ, Information for candidates: Coursework assessments, Instructions for conducting coursework and Instructions for conducting non-examination assessments (all effective 1 September 2026), and About us, read 2 October 2026.
• OCR, AI guidance for NEA components and A Level History A coursework guide, read 2 October 2026.
• Pearson, EPQ AI evidence template, read 2 October 2026.
• AQA, Extended Project 2026/27 candidate record form, read 2 October 2026.