Artificial intelligence may be moving beyond research assistance and into the role of legal tutor, according to a new study from Stanford Law School that produced a surprising outcome: law professors consistently favored AI-generated responses over those written by their academic peers.
The research set out to test whether AI systems could effectively answer the kinds of questions first-year law students typically ask during office-hour discussions. Professors from 14 law schools across the United States compiled 40 questions commonly raised by students studying contract law.
Faculty members first drafted their own answers. Researchers then tasked two AI platforms—Google’s Gemini 2.5 Pro and NotebookLM—with responding to the same questions.
The resulting answers were presented to participating professors without revealing their source. When asked to choose which responses would be most useful to students, the professors selected the AI-generated versions roughly three-quarters of the time.
Even more striking, the performance of the AI systems matched that of the highest-rated professor in the study.
Researchers noted that the questions were not limited to basic legal concepts or straightforward textbook explanations. Many required nuanced reasoning and detailed analysis, making the results more significant than a simple test of factual recall.
The findings arrive as law schools continue debating how best to integrate artificial intelligence into legal education. Previous studies have already shown that advanced AI models can perform strongly on law school exams, achieve high academic marks, and even pass bar-exam-style assessments.
While concerns about misuse remain, some institutions have begun requiring students to learn how to work with AI tools. Others have moved in the opposite direction, tightening restrictions on their use in coursework.
The Stanford study points to another potential role for the technology: providing students with immediate access to academic support. Instead of waiting for office hours or relying on classmates for clarification, students could use AI systems to obtain explanations whenever questions arise.
Researchers also found that harmful or misleading responses were relatively uncommon among the AI-generated answers. Fewer than 4% were judged to be detrimental to student learning, compared with 12% of the answers written by professors.
The authors concluded that AI-powered tutoring systems could serve as a valuable complement to classroom instruction, expanding access to expert-level guidance while offering students reliable, on-demand assistance.
