Professor Thibault Schrepel of Vrije Universiteit Amsterdam conducted a two-year study on the impact of AI tools in legal education. The experiment involved 66 students in 2024 and 164 participants in 2025. Students were divided into three groups: those banned from using ChatGPT, those using it without guidance, and those trained in prompt engineering.
The study found that students without access to AI consistently performed the worst. While the group with structured training initially outperformed others in 2024, this performance gap narrowed significantly by 2025 as general familiarity with AI tools increased among all students. The unguided group also outperformed the no-AI group, suggesting that students learn to identify AI errors through practical use.
Schrepel argues that universities should move away from blanket AI bans and instead focus on teaching ethical use and legal responsibility. He suggests that traditional assignments, such as literature reviews, are becoming obsolete as AI can complete them in minutes. Instead, he advocates for practice-oriented or empirical coursework where AI integration is evaluated.
The research highlights a conflict with institutions like UC Berkeley Law, which maintains strict AI bans for graded work to preserve core thinking skills. Schrepel acknowledges limitations in his study, including a small sample size and the tech-savvy nature of his students. He notes that other research shows AI often boosts homework grades while potentially hindering performance on proctored exams.
