Asked to Explain Your Own Paper: Oral Defence Checklist and What It Proves

You are being asked to sit an unscheduled oral assessment about your own document, and what you say may be written down. The Office of the Independent Adjudicator, the student complaints scheme for England and Wales, published six case summaries on AI and academic misconduct in July 2025, and three of them involved a viva. In one, the ombudsman upheld the student's complaint, and one of its reasons was that the viva "explored the student's knowledge of the subject but did not give the student an opportunity to also explain how they had worked on the assignment". So an oral defence can test two different things, subject knowledge and process. You do not get to choose which one you are asked about, so prepare for both.

The record of the meeting matters afterwards too. Another of the three cases, covered below, is about a provider whose account of a viva did not match the transcript.

If you are studying in a second language, the section on explaining badly has published IELTS figures on speaking and writing that are worth knowing before you go in.

The six summaries sit under "AI and academic misconduct" on the OIA's site, dated July 2025. Three describe a viva: CS072502, CS072503 and CS072504. Of the other three, one was decided on written evidence, one was settled, and one was a student complaining that their feedback looked AI-generated.

Put briefly, the decision in the first rested on the viva alone, in the second the viva was one piece of evidence among several, and the third is the transcript case mentioned above. We walk through the same cases from the appeal side, including the one that was not about a viva, in Falsely flagged for AI: how to appeal, and what evidence your university will accept. This article stays with the meeting itself.

The CS072502 finding is useful because it splits the thing in half.

Subject questions are about the content. What is the argument. Why these sources. Where this figure came from. What that sentence means. They test whether you understand the paper.

Process questions are about the making. When you started, what you did first, what changed between drafts, what you used. They test whether the document has a history you can narrate.

The ombudsman treated a viva that asked only the first kind as incomplete. Not wrong, incomplete. So the process half is not a side topic you have to squeeze in. It is part of what a fair oral defence covers, and if nobody asks, you can offer it.

At least one university publishes the process questions it uses, which is more useful than any list we could write. The University of South Carolina's academic integrity office publishes these verbatim:

"Did you use any online assistance to complete your assignment?" · "How long did you work on this assignment? Was it in one sitting or over the course of several days?" · "Did you use any software to edit your paper?" · "Could you walk me through your process so we can determine if a tool you used may have been powered by artificial intelligence?"

Turnitin publishes a similar set for instructors, and two of its four opening lines are about process rather than about AI. That page is covered line by line in what your instructor is told to do when your AI percentage is high.

None of the published sets we found includes subject questions. Those come from whoever read your paper, and they will be specific to it. So that half cannot be prepared as a script. The preparation is the paper itself.

1. Re-read your own paper, out loud, the day before. You may not have opened it in weeks. In CS072502 the student's appeal said that "a significant period of time had elapsed between the original submission and the viva which had affected their recall". That can happen to someone who wrote every word.
2. Be able to say where each source came from. Not a summary of it. Where you found it: the reading list, a reference in another paper, a database search, a supervisor's suggestion. That answer rests on searching you actually did, which is why it comes easily to someone who did it.
3. Know why your flagged paragraphs, and your three weakest ones, are there. Those are the passages most likely to be asked about, and "I struggled with that section and here is what I was trying to do" is an answer only the author can give.
4. Have the process ready as a timeline, with something behind it. Start date, order of writing, what changed and why, what you cut. Then the material that backs it: drafts, version history, notes, emails. How to keep version history in Word, Google Docs and Overleaf covers where that record lives in each tool, including the Word case where it does not exist at all unless the file was in OneDrive.
5. Ask, in writing, what the meeting is and what evidence it will use. Whether it is an informal conversation or a step in a formal procedure, whether it is recorded, whether you can bring somebody, and whether you can see the evidence beforehand. In CS072502 the ombudsman's first criticism was that "none of the evidence relied upon by the provider was shared with the student ahead of the viva", so asking for it lines up with what the provider was expected to do.
6. Straight afterwards, write down what you were asked and what you said. Do it the same day, while it is exact. The next section is why.

Once the meeting is over, what is left of it is somebody's written account.

In CS072504 the panel worked from "a full transcript of the student's viva". The provider said the student had admitted a particular use of AI during the viva. The ombudsman found that the panel "misrepresented what the student had said during the viva", and the transcript "recorded no admission" of that use. The difference could be seen because a full transcript existed.

Where a transcript or recording exists, it can protect you, so it is worth asking whether one is being made. Where it does not, the record is somebody's notes, and their summary is what a later panel reads.

That is why your own note matters. It does not prove what was said, but a detailed account written the same day, sent to yourself or to a student adviser, is far more useful to have than your memory of the meeting three weeks later.

The University of York's academic misconduct policy for 2025/26 sets out what a formal interview looks like at one institution, and its procedural detail is worth reading even if you are somewhere else. Students must be given "sufficient time before the interview to seek advice or to arrange to be accompanied (at least 5 working days)", any interview "must include at least two members of the StAMP Investigatory Panel, including the Chair", and "The Panel will only base its decision on evidence the student has had sight of in advance and had the opportunity to respond to." Your own institution's procedure is the document that governs you, and finding it is part of preparing.

Speaking and writing are assessed as separate abilities by the organisations whose job it is to assess them, and the gap between the two is measured and published.

IELTS reports four band scores, one each for Listening, Reading, Writing and Speaking, and it publishes population means. Its test taker performance file for 2024–2025, downloadable from ielts.org, has a sheet named "AC mean by first languages": 40 rows, 39 first languages plus one labelled "Other", each giving mean band scores. Comparing the "Writing Band Score" and "Speaking Band Score" columns on the unrounded values, mean Speaking is above mean Writing in 35 of the 40 rows, "Other" included.

The five where it runs the other way are the ones to notice.

Differences are computed from the unrounded values in the file, so they can differ by 0.01 from subtracting the rounded columns. The same comparison on the General Training sheet in the same file also gives 35 of 40 rows with Speaking higher, and four of its five exceptions are the same languages, so the pattern is not an artefact of one sheet.

Two limits on what the table can mean. These are population means for people sitting a language test, not predictions about any individual, and we found nothing linking an IELTS Speaking band to how someone does in a viva. And IELTS's statistics page says the language tables show "the top nationalities and languages listed alphabetically", so it is not every language.

What it does establish is narrower, and still useful. For test takers of those five first languages, the average Writing band was higher than the average Speaking band in 2024–2025. That says nothing about you, but it means a weaker spoken performance than written one is a documented pattern for some groups, not an excuse invented in the room. If an oral defence goes badly, the thing that went badly is the oral performance.

Other pressures sit on top of that. It is unrehearsed, it can come weeks after submission, and you are there because something has been alleged. Western Washington University's guidance for its own instructors lists non-native English speakers and students with accommodations among the situations where it tells instructors to recognise the potential for false positives. That guidance is about the detector. It seems reasonable to carry the same caution one step further, into the room.

If a second language or a registered accommodation applies to you, raise it, but raise it early and in writing, as a request rather than as an answer to a question about the paper: ask what adjustments the procedure allows, and ask that written material be considered alongside whatever happens verbally. In CS072504 the ombudsman faulted a provider for not considering whether AI detection "might be less reliable for non-native English speakers", and in CS072502 for not letting the student explain how they had worked.

That last row is the counterweight to the idea that an oral defence is a way to clear your name. A viva records what you say about how the work was made. That helps when the account is accurate and backed by material, and it cannot make an account true.

Requesting an oral defence can look like the obvious way to prove you wrote something. At least one university's own procedure tells its students not to assume that step will help them.

The University of York's 2025/26 policy says that a student may respond to the panel with a written statement, by requesting an interview, or both, and adds a parenthesis: students "should be made aware that there is no inherent benefit to an interview". That is York's guidance to its own students about its own process, not a general rule. But it is a real institution telling students that the meeting is not automatically to their advantage.

The answer differs from person to person. An oral defence lets you answer a specific doubt in real time, and a misunderstanding can be cleared up in one sentence instead of three emails. On the other side, an unprepared answer becomes a recorded one, and speaking under pressure is the skill being sampled. And if there is a written route, a written statement is something you can draft, check and keep a copy of.

If the meeting is compulsory, the question does not arise and the checklist above is the whole job. If it is optional, ask what the written alternative is before deciding, and ask your students' union or your institution's student advice service. They are more likely than we are to know how these run where you study.

HumanPen, which we make, is a tool that rewrites documents. It has no role in a viva or in the meeting around one.

The one adjacent thing worth saying is about timing, and it argues against buying anything for now. Rewriting a paper before a meeting about that paper changes what you are being asked about, and if a version is already sitting with your department, that version is the one to discuss. If a revision is agreed as the outcome afterwards, keep it small. Ours only touches what you fence off first, and there are two ways to build that fence: pick the passages yourself, or bring in a detection report so the passages it highlighted become the selection. A whole paragraph is the finest resolution on offer, so half a paragraph turns into all of it, and that expansion has to be accepted by you before the run begins. Eligible passages can be re-run at no charge. The output stays an editable document, designed to preserve citations, tables and layout. Starting from a report is described in Rewriting only the paragraphs a Turnitin report flagged.

What will they ask me? The University of South Carolina publishes its process questions, and Turnitin publishes conversation openers for instructors. Most of them are about how the work was done: how long it took, whether it was one sitting or several, what software was used, walk me through your process. Turnitin's set also proposes going through the highlighted passages together. Subject questions come from your paper, and none of the published sets we found includes them.

Can I be made to attend? It depends on your institution's procedure, and some do have the power. The University of York's 2025/26 policy says: "Even if the student does not request an interview, the panel may require the student to attend one." Find your own institution's document rather than assuming; it is the one that governs your case.

Can I bring someone with me? At some institutions, yes, sometimes with notice requirements attached. York's 2025/26 policy gives students "at least 5 working days" before an interview to seek advice or arrange to be accompanied, and recommends contacting York SU advice and support. Ask when the meeting is proposed rather than on the day.

I froze and could not explain my own argument. Is it over? Not necessarily. A viva is one input, and the table above sets out what it can and cannot carry on its own. Write down what happened the same day, and put the process material in writing: drafts, notes, timeline, sources.

English is not my first language and I am worried about the meeting. For some groups the gap is measured. On IELTS Academic population means for 2024–2025, mean Speaking is above mean Writing in 35 of the 40 rows (39 named first languages plus "Other"), but for Chinese, Japanese, Korean, Vietnamese and Punjabi first-language test takers the mean Writing band is the higher one. The practical step is the one in the section on explaining badly: raise it early and in writing, and get your process account on paper before the meeting so the spoken performance is not the only record.

Should I quote the ombudsman cases at my university? They are published case summaries about other providers, and they do not bind yours. They are useful for deciding what to prepare and what to ask for. What belongs in the room is your paper and your process.