Key points:
In 2023, Catherine Hartmann, a professor at the University of Wyoming, gave students in a history of meditation course a deeply personal assignment: try a contemplative practice and write a reflective analysis on the experience.
One student turned to AI instead. The evidence was impossible to miss – the student had accidentally left the copy-pasted AI prompt sitting right at the top of the submitted paper.
For Hartmann, the issue extended far beyond one student using a shortcut. Generative AI in education was quietly fracturing the foundational contract between teacher, student, and classroom assessment.
Take-home writing, long considered evidence that students had read, analyzed and understood the material, could now be generated or substantially shaped by AI in seconds. Teachers risked shifting from educators into digital detectives, scrutinizing student submissions for signs of machine authorship rather than evaluating genuine human thought.
To move forward, Hartmann looked backward: she brought back the oral exam.
She redesigned her upper-level humanities course around oral assessment. However, rather than simply replacing the final paper with an unscripted conversation and hoping students could perform, she used backward design to restructure the entire course.
Throughout the semester, students regularly discussed complex ideas, responded to questions, and practiced articulating their thinking. The final oral exam became the natural culmination of a way of thinking and communicating they had practiced all term.
There is a striking paradox at play here. At the exact moment education is grappling with its newest, most disruptive technology, an assessment practice centuries older than the computer feels remarkably modern.
Generative AI hasn’t made student assessment obsolete. Instead, it is forcing educators to reconsider a fundamental question: What actually counts as evidence that learning has taken place? The shift is not confined to the humanities. At the University of Pennsylvania, mathematics professor Robin Pemantle has students work through calculus problems at the board while explaining their reasoning aloud, allowing him to assess not only whether they reach an answer, but how they think their way toward it.
The renewed interest in oral exams is not simply about catching students who use AI. If catching cheaters becomes the primary goal, education risks replacing one form of surveillance with another.
The more important opportunity is to reconsider what assessment is supposed to reveal. A finished essay gives us an outcome. An oral exchange gives us access to the thinking behind it.
When an educator asks a student to explain why they chose a particular method, defend an interpretation, or adapt an answer when a condition changes, assessment becomes dynamic. Teachers can hear misconceptions as they emerge, probe partial understanding, and distinguish between memorized reproduction and knowledge a student can actually use.
AI doesn’t need to be completely banished from this process. A student might use AI tools to brainstorm, test arguments, or prepare for an assessment. The crucial question shifts from “Did AI touch this work?” to “Can the student take intellectual ownership of what they are presenting?”
Safeguarding equity in spoken assessment
Bringing back oral assessment requires intentional care. Not every student arrives equally prepared to think aloud under pressure.
For multilingual learners in particular, hesitation, accent, vocabulary retrieval, or processing time in an additional language must not be mistaken for a lack of subject knowledge. If the goal is to assess mathematical reasoning, historical analysis, or scientific understanding, language must serve as a bridge—not an unintended barrier.
Hartmann’s approach at Wyoming offers an important lesson: oral assessment should not begin on exam day.
Students need scaffolded, low-stakes opportunities to discuss, explain, rehearse, and respond throughout instruction. The final evaluation should reflect a form of classroom participation they already know well.
An old assessment for a new era
The return of the oral exam does not mean abandoning essays, research papers, or written projects. Nor does it mean attempting to build a bulletproof assessment that AI can never touch, a race education is unlikely to win.
Instead, generative AI offers educators an invitation to broaden what counts as evidence of learning. A written product shows us what a student produced. A conversation can reveal what the student understands, how they arrived there, and what happens when their thinking is challenged.
Perhaps that is why the oral exam suddenly feels new again. In an age when machines can instantly generate answers, assessment can no longer just be about collecting artifacts. It must become about encountering the human mind behind them.
