Verbal authentication of coursework: A JCQ-aligned protocol for short student discussions

 A teacher reads a piece of coursework and pauses. The conclusion is unusually polished. The vocabulary feels unfamiliar. Perhaps there is an innocent explanation, or perhaps the work does not reflect the pupil’s independent effort.

That judgement matters. A pupil who genuinely understands the subject may struggle to explain themselves under pressure. A teacher may have only fragments of evidence from lessons and drafts. A parent will reasonably want any consequential decision to be fair. There is no perfectly tidy answer.

 A short verbal discussion can help. In its April 2025 guidance, JCQ recommends considering a discussion with the student to establish whether they understand their work and whether it reflects independent effort. It is one source of evidence not a universal requirement or a prescribed test. 

 Used well, the conversation gives a pupil an opportunity to reveal the thinking behind the page. Used badly, it becomes an improvised oral examination in which confidence is mistaken for honesty. The difference lies in purpose, questioning and what happens next.

 What the conversation is for 

 Authentication is broader than spotting suspicious prose. Teachers draw on their knowledge of the candidate’s usual standard, supervised work, drafts, research and regular discussion. JCQ’s instructions for conducting coursework describe familiarity, discussion and direct supervision as parts of this process. However, these instructions apply specifically to ELC qualifications, Project qualifications and CCEA GCE unitised AS and A-level qualifications. Separate JCQ instructions cover relevant GCE and GCSE non-examination assessments in England and Wales. Subject-specific and awarding-organisation rules still govern each component. 

 A verbal check should explore whether the student can explain a specific part of the submission, connect it to a plausible process and work with the reasoning it contains. It should not award extra marks, improve the submitted answer or serve as automatic proof of malpractice. Nor should it be triggered simply because a pupil is quiet, has English as an additional language, has SEND, performs better in coursework than in examinations or uses unexpectedly sophisticated vocabulary. There should be a question to resolve: perhaps a gap between drafts and the final version, limited opportunity to observe the work developing, uncertainty about an individual contribution, or missing records of declared AI use. The same caution applies to automated detection. JCQ says detector results vary with the tool, model, editing, prompts and English language competency. They should be considered alongside all available evidence, not treated as a verdict 

 Four to six minutes is often enough for a focused exchange, although JCQ does not prescribe a duration. The TALK structure keeps the conversation anchored to the work rather than the pupil’s demeanour. 

 1. Tell the student why you are asking. Explain that this is an authentication check, not a newly marked assessment. Invite them to ask for a question to be repeated or rephrased.

2. Ask about the artefact and its development. Choose one or two precise points. Ask what the student was trying to do, what changed between drafts, or where an idea appears in their notes or source record.

3. Link the work to reasoning and evidence. Ask why a source supports the conclusion, why one method was chosen over another, or what limitation had to be managed. If AI use has been declared, ask what the tool was asked to do, what material was used and how it was checked. For covered coursework and non examination assessments, JCQ requires candidates using AI to retain prompts and generated material in a non editable format and explain how it was used. A conversation supplements that record. It does not replace it. 

 4. Kick the reasoning one step further. Offer a modest variation: “If you removed this source, how would the conclusion change?” or “If you repeated the investigation with more sites but used the same selection method, which problem might remain?” The pupil is not being asked to produce another polished answer. The question tests whether they can work with the reasoning already present. 

 Consider a geography investigation whose final conclusion contains a far more sophisticated account of sampling bias than the supervised planning work. “This does not sound like you. Did you use AI?” starts with an accusation and produces little reliable evidence. A better opening is: “I want to understand how this conclusion developed. Show me the result that made you question the reliability of the data.” The teacher might then ask what sampling bias means in this investigation, which source informed the paragraph and what would happen if the number of sites increased without changing the selection method. Suppose the student explains the concept but discloses using an AI tool to rewrite the paragraph without retaining the prompts or output. The teacher now has evidence requiring review. The conversation has not, by itself, established malpractice. 

 Read the evidence, not the performance 

 Fluency does not prove authorship. A student may have rehearsed an explanation or received extensive help. Hesitation does not prove that work is inauthentic. Anxiety, memory, language proficiency and communication needs all affect spoken responses. Research on oral assessment supports careful design, consistent examiner behaviour and clear rules for prompting. It also identifies anxiety and possible bias as threats to validity. Much of this literature comes from higher and professional education, so it does not validate any particular school authentication protocol. It does, however, reinforce the need for restraint 

The teacher should consider three things together: whether the pupil can explain central ideas in the submission, whether their account of producing it fits the available drafts and records, and whether the conversation is broadly consistent with supervised and previous work. No single awkward answer should decide the case.

 Record enough to support the judgement 

 Notes should capture evidence rather than impressions such as “seemed nervous” or “looked guilty”. Record the work discussed, the principal questions, a factual summary of the responses, any drafts or source records checked, declared assistance or AI use, and any adjustments provided. Also record whether the candidate had signed the authentication declaration, because that can change the procedural route.