Oral exams have become one of the most widely discussed responses to generative AI in higher education. If a student can talk you through their reasoning, answer follow-up questions and explain why they made the choices they did, you learn something that a written submission alone can no longer reliably tell you. Earlier this year I wrote that oral exams are back – and that the real challenge is operational rather than pedagogical. That still holds.
What has changed since then is the market. More and more assessment platforms now list “oral exam” among their features. That is a good thing – it shows the sector has recognised the need. But it also makes it worth asking a simple question before choosing a solution: what exactly is being digitised?
Because an oral exam is not a question you put in a test. It is a process that starts long before the student enters the room and ends long after they leave it. And the format institutions are now reaching for most – a written assignment followed by an oral defence – only works if that process holds together from start to finish.
Think about what actually happens around a single oral exam. Before it, students need a time slot and a room, and sometimes preparation time with a question drawn in advance. Examiners need to know who they are seeing and when, and – crucially – they need to have read what the student submitted. If there is a co-examiner or an external examiner, they need the same overview. Students with special arrangements may need extra preparation time or a different setting.
Then comes the conversation itself, which is the part everyone pictures. Afterwards, the grade has to be agreed and registered, preferably on the same day. The examiners’ notes and the assessment criteria they used need to be kept somewhere they can be found again. And if the student later disputes the result, the institution must be able to show how the decision was made.
If a platform treats an oral exam as a question type – a test containing nothing but “oral” items, with a button to start a timer and a field to enter a grade – it has digitised the stopwatch, not the exam. Everything else is left to email, shared calendars and spreadsheets. That may work for a single course. It rarely survives an exam period.
Very few institutions are replacing written assessment with oral exams outright. What they are doing instead is combining the two. Students write a project, a report or an essay at home, and then defend it in a conversation with their examiners. I have argued before that combining assessment formats is one of the most robust ways to diversify assessment in the age of AI, and this is the clearest example.
The logic is sound. The written part lets students work in depth, over time, with the resources they would use in real life – which may well include AI. The oral part checks that the understanding behind the text actually belongs to the student. Neither part does the job on its own. Together, they give a much fuller picture.
But this is also exactly where the demands on the platform increase. A combined exam is not two separate exams that happen to share a student. It is one assessment in two stages, and the second stage depends on the first.
An oral defence gets its value from being specific. The examiner asks about this paragraph, this choice of method, this conclusion. “Why did you leave out this source?” is a question that tests ownership. “Tell me about your topic” is not. That means the link between the written submission and the oral exam is not a nice-to-have. It is the whole point.
In practice, that link has three parts:
Access. The examiners – including any co-examiner or external examiner – must have the student’s submission in front of them, both when preparing and during the conversation itself.
Eligibility. The written part should be a condition for the oral. If a student has not submitted, the system should know that before a time slot is used.
Coherence. The two stages must come together as one assessment with one documented outcome – not two records in two systems that someone has to reconcile by hand afterwards.
A standalone oral module, however well designed, cannot provide this on its own. If the written submission lives somewhere else, the examiner has to fetch it from somewhere else, and the institution has to stitch the result together itself. That is precisely the manual work that makes oral exams expensive – and precisely the place where errors creep in when hundreds of students are assessed in the same few weeks.
At Danish universities, oral exams are not a new response to AI. They have been a core part of the assessment tradition for a very long time, often with an external examiner present to safeguard the standard. That is the environment WISEflow was built in, and it has shaped how we support oral assessment – not as a feature added recently, but as something that has been run, adjusted and refined through many exam seasons.
In WISEflow, the oral exam has its own flow, FLOWoral, and it can be placed in a sequence of flows after a written assignment. The examiners can read the written submission in the first flow, the system checks that the student has submitted before they move on, and the oral exam follows in the next flow with its own schedule. Time slots can be allocated automatically, with room information and preparation time where the exam requires it. Co-examiners and external examiners have their own roles, special arrangements are handled as part of the planning, and examiners can make notes and work with assessment criteria during the exam before registering the grade.
Most oral exams still take place face to face, and we think that is right. Where an institution wants to run them remotely, a link to the video tool of its choice – Microsoft Teams, for example – can be added to the exam, so that the institution’s existing infrastructure and policies, including any decision about recording, stay in its own hands.
None of this is spectacular on its own. Its value lies in the fact that it all sits in one place, built around the way oral exams are actually run.
If your institution is expanding its use of oral exams, these are the questions I would put to any platform – including ours:
1. CAN THE ORAL EXAM FOLLOW A WRITTEN SUBMISSION IN THE SAME ASSESSMENT?
Or is it a separate test that has to be linked to the written part by hand? If oral exams can only contain “oral” items, the combined format will live outside the system.
2. DO EXAMINERS HAVE THE STUDENT’S WRITTEN WORK IN FRONT OF THEM?
Both when preparing and during the exam – and does that apply to co-examiners and external examiners as well?
3. WHO BUILDS THE SCHEDULE?
Are time slots, rooms and preparation time handled in the platform, or exported to a spreadsheet? And what happens when a student is ill on the day?
4. HOW ARE SPECIAL ARRANGEMENTS HANDLED?
Extra preparation time or a different setting should be part of the planning, not a manual exception. I have written more about this in Rethinking special arrangements.
5. WHAT CAN YOU SHOW IF A STUDENT APPEALS?
Notes, criteria, who was present and when the grade was registered. Oral exams are the hardest format to document, which is why defensible marking matters more here than anywhere.
The renewed interest in oral exams is one of the more encouraging developments in the debate about AI and assessment. Unlike detection tools or a wholesale return to the exam hall, it moves assessment forward rather than back – towards formats that test understanding rather than output.
But the pedagogical idea only becomes institutional reality if the process around it can carry the load. An institution that adopts oral exams with a tool that covers only the conversation will find that the cost moves to administrators and examiners instead. And when that cost becomes too high, oral exams quietly disappear again – not because they did not work, but because they could not be run.
Don’t ask whether a platform supports oral exams. Ask whether it supports the whole process – before, during and after the conversation, and together with the written work being defended. That is the difference between an oral feature and an oral exam.