For centuries, at Oxford University, the viva voce was the examination. A student who had submitted a thesis would be placed before a panel of examiners and required to defend it verbally — to explain, justify, and argue for every claim, in real time, under questioning. Viva voce: Latin for "by living voice." No notes. No preparation beyond the work itself. Just the mind, the mouth, and the examiner.

The written exam eventually displaced this tradition in most universities, for good practical reasons: it scaled, it was reproducible, it was quantifiable, it was efficient. One professor could evaluate 200 students through a written test. The viva voce required two examiners, an hour per student, and a room.

And then came AI.

Across the United States, universities are returning to oral exams and in-person assessments. Professors report that students submit perfect assignments but struggle to explain the material when questioned directly.

The pattern is now familiar on every campus that uses AI-detection software, and on many that don't. Students submit flawless essays. When questioned about the content, they struggle to explain it. The essay measured nothing except the student's ability to prompt an AI. The exam had become a test of AI, not of the student.

Chris Schaffer, a biomedical engineering professor at Cornell, introduced what he calls an "oral defense" and put it plainly: "You won't be able to AI your way through an oral exam."

What AI cannot do — what no current AI system can fully substitute for — is speak for you in real time, in a room, under direct questioning by an examiner who is watching your face and following up on every hesitation. The oral exam is the one format where the intelligence being measured is unambiguously the student's own.

The implications extend far beyond academic integrity. The oral exam is also a fundamentally different measurement instrument. It assesses not only whether a student knows something but whether they can articulate it, defend it under challenge, adjust their explanation in response to confusion, and think out loud in real time. These are the skills that employers consistently identify as most valuable and most absent in recent graduates. They are also the skills that AI is making students less likely to practice.

At NYU, a remarkable experiment is already underway. The university piloted an AI-powered oral examination for a 36-student undergraduate course. A human-administered equivalent would have cost $750. The AI-driven version cost $15 total — approximately 42 cents per student. For the first time in the viva voce's 2,000-year history, the oral exam is scalable to large undergraduate courses.

The 42-cent oral exam. A figure that makes the viva voce scalable for the first time in its 2,000-year history.

Oxford never stopped using the viva voce. It has been using it for doctoral examinations since the university was founded. What the AI era has revealed is that Oxford's most ancient assessment format was also its most future-proof.

The oral exam is not a nostalgic retreat. It is the answer to a problem that technology created and technology — in a different form — is now helping to solve.

(Sources: The Conversation / Fine Day Radio / San Diego Today / Wedbush Financial / The Chronicle of Higher Education / ArXiv — 2025–2026)

Seoul 🇰🇷 — South Korea has the world's most intense written examination culture. The suneung — the College Scholastic Ability Test taken every November by hundreds of thousands of students — is a single written exam that determines, to an extraordinary degree, the trajectory of a Korean life. The educational infrastructure built around it — the hagwon network, the years of test preparation, the social pressure organized around a single performance on a single day — is built entirely around written answers to fixed questions. AI has not yet disrupted the suneung because the exam is conducted on paper, in a strictly controlled environment, without internet access. But the AI disruption that Oxford is responding to with the viva voce will reach the Korean university system — the assignments, the essays, the reports that make up the rest of a Korean student's academic life. When it does, the oral exam will encounter a specific resistance: Korean educational culture does not train students to speak their knowledge. It trains them to write it. The viva voce requires not only knowledge but the confidence to perform it aloud. That is a different skill set. Seoul's educators know this. The conversation has not yet become a policy.

New York 🇺🇸 — New York's universities are at the front line of the oral exam revival. At Cornell, Chris Schaffer is using oral defenses in biomedical engineering. At Penn, the executive director of the Center for Teaching and Learning describes what is happening as "a massive shift toward in-person assessments." At NYU, the 42-cent AI oral exam pilot has produced results that are being watched across American higher education. New York's specific contribution to this story is scale: the city's universities enroll hundreds of thousands of students. If the oral exam revival is to work at New York's scale — not just in doctoral programmes but in undergraduate courses of 200 students — it requires exactly the kind of AI-assisted oral examination that NYU is testing. The most ancient form of assessment and the most advanced form of AI are meeting in New York to solve the same problem.

Singapore 🇸🇬 — Singapore's higher education system is one of the most internationally connected in the world. NUS and NTU consistently rank in the global top 20. The country's approach to AI in education has been characteristically systematic: the Ministry of Education has been developing AI literacy frameworks, responsible use guidelines, and assessment reform recommendations since ChatGPT's emergence. What Singapore has not yet produced — and what Oxford's viva voce tradition offers — is a deep cultural comfort with spoken academic performance. Singapore's educational culture, like Korea's, has been built around written examinations and quantifiable outputs. The oral exam introduces something that systematic frameworks find difficult to measure: the quality of thought expressed spontaneously under challenge. Singapore is well-positioned to build the AI infrastructure for oral assessment at scale. Whether it can also rebuild the pedagogical culture that makes students comfortable performing their knowledge out loud is a different question.