Schools return to oral exams as generative AI renders written assignments unreliable

Educational institutions are increasingly reviving oral examinations to verify student knowledge and combat the rise of generative AI cheating. This pedagogical shift marks a return to interpersonal evaluation as traditional written assessments become easier to simulate.

In Denmark, the government has introduced measures requiring upper-secondary students to verbally defend written assignments completed at home. To further secure the process, Danish schools have been instructed to implement screen-monitoring tools and firewalls during written examinations to prevent the use of AI. These steps are being taken because AI detection software is often unreliable and prone to false positives, making it difficult for educators to prove academic dishonesty through software alone.

The collapse of the take-home essay

For years, education shifted toward scalable, written formats and take-home essays to increase efficiency. However, the proliferation of Large Language Models (LLMs) has rendered these formats vulnerable to systemic cheating. When the primary evidence of learning—the written paper—can be simulated by an algorithm, the traditional trust between student and instructor is eroded. This has forced a return to proctored, in-person assessments that were previously phased out in favor of more flexible digital submissions.

The equity trade-off

While oral exams provide a real-time check of authenticity, they introduce significant concerns regarding fairness and scalability. Critics argue that these high-stress evaluations can disadvantage neurodivergent students and those struggling with high anxiety.

There is a growing concern regarding a "salesperson" versus "deep processor" dynamic. In this scenario, students with high social confidence and charisma may be able to mask gaps in their knowledge, while those who require more time to synthesize complex answers are unfairly penalized. Because these exams are difficult to scale and administer fairly, they are not viewed as a simple fix for the integrity crisis.

Redefining academic evidence

This trend highlights a fundamental crisis in how academic achievement is measured. Institutions now face a choice: they must either redefine what constitutes "learning" in an AI-augmented world or commit to labor-intensive, interpersonal evaluations that may introduce new forms of grading bias.

As more countries follow Denmark's lead, the focus will likely shift toward hybrid models. Educators are now watching to see if verbal defenses can be standardized to protect anxious students while still providing the verification that software cannot.