Law Professors Are Turning to Oral Exams as AI Makes Written Tests Nearly Unverifiable

The Cheating Problem Has Outpaced the Proctoring Tools

When ChatGPT launched in late 2022, law professors had a few years of relative calm. Proctored exams on locked-down laptops felt like a workable defense. That window has closed.

According to a piece published by Reason, professors are now watching YouTube tutorials where students demonstrate how to defeat webcam proctoring: applying vaseline or another agent to obscure the lens, photographing exam questions with a phone or smart glasses, and feeding them to an AI for a polished answer. The professor who wrote the piece, Josh Blackman of South Texas College of Law Houston, notes that take-home work is even more compromised. A student who chooses NOT to use AI is, practically speaking, competing at a disadvantage.

The integrity problem isn't just about cheaters. It's about the structural incentive AI creates for everyone.

What Schools Are Actually Doing About It

Blackman points to a faculty memo from Bobby Chesney, Dean of the University of Texas School of Law, as one of the more serious institutional responses on record. According to Blackman's account of the Chesney memo, UT Law faculty have seen "a surge of interest in assessments involving class participation and live presentations — even oral exams in some cases."

Oral exams aren't new. They're the oldest form of academic evaluation in Western education. The doctoral defense, the Socratic method, and the bar exam's practical component all have roots in live interrogation of knowledge. What's new is that written exams, which displaced oral ones largely for reasons of scale and standardization, are now the format with the integrity problem.

Blackman says his own tentative plan for Constitutional Law next spring is to switch to an in-person oral midterm.

The Legitimate Concern About Oral Exams

Critics of oral exams raise real objections that deserve a fair hearing. Students with anxiety disorders, language barriers, or neurodivergent processing styles may perform significantly worse under live verbal pressure than their actual mastery of the material would warrant. A written exam, whatever its AI vulnerabilities, at least gives every student the same controlled time and format. Oral exams introduce subjectivity: a professor who unconsciously favors confident speakers, or students from backgrounds where articulate verbal performance is trained from childhood, may inadvertently bake in a different kind of unfairness.

Those are valid structural concerns. They don't make oral exams wrong, but they mean the format requires deliberate design, not just a default switch because written tests got hard to police.

The Midterm Problem Isn't Just About AI

Blackman is candid that the oral exam idea also responds to longstanding frustrations with midterms that predate AI entirely. He describes a recurring problem: by the time a written midterm is graded and returned, it's often too late for struggling students to course-correct. He used to proctor midterms right before spring break, grade through the break, and return exams when the semester resumed. A schedule like that no longer works given family obligations.

He also notes that midterm performance in Constitutional Law didn't always predict final exam performance. Some students hadn't hit their stride by week seven. That's an honest admission from someone thinking through assessment design.

The "Use It Responsibly" Answer Isn't an Answer

Some faculty have moved in the opposite direction, embracing AI and asking students to document their prompts and reasoning. Blackman is skeptical of this approach, calling himself an "AI luddite" who writes without generative tools. He doesn't dismiss the embrace-AI camp entirely, but his critique is that "use it responsibly" is undefined guidance that puts students in an ambiguous position about what's actually being evaluated.

If the goal of a law school exam is to verify that a specific student can analyze a legal problem, then an exam that allows AI-assisted answers is measuring something different. Law schools should say clearly what that something is, rather than treating vague "responsible use" standards as a solution.

What Remains Unresolved

The practical question Blackman's piece leaves open is one the broader academic community hasn't answered: how do you scale oral exams across a full course load without either destroying faculty time or cutting so many corners that the format loses its diagnostic value? A single oral midterm for a ConLaw section is feasible. An oral final for every course at a large law school is a logistical problem no one has yet solved at scale.

Whether law schools converge on oral assessments, AI-integrated assignments with transparent expectations, or some hybrid format will likely be settled one faculty decision at a time over the next few years without a coordinated institutional standard in sight, as of June 22, 2026.