Stanford Study Finds AI Outperforms Law Professors in 75 Per Cent of Cases

A Stanford Law School study has found that law professors preferred answers generated by artificial intelligence over those written by fellow academics in 75 per cent of cases, suggesting the technology may already be capable of supporting legal reasoning and student learning.

Professors from 14 US law schools compiled 40 questions modelled on those commonly raised by first-year contracts students during office hours. The researchers then produced written answers to the questions and compared them with responses generated by two AI systems — Google’s Gemini 2.5 Pro and NotebookLM.

The same professors then blindly assessed the answers in head-to-head comparisons, consistently selecting the AI-generated responses as more beneficial to students in three-quarters of cases. The study found that the AI systems performed on a par with the highest-rated professor-written responses.

“We were frankly surprised by the magnitude of the results,” said lead researcher and Stanford Law professor Julian Nyarko in comments published on the university’s website. “These weren’t just simple questions with obvious answers.”

The findings come as law schools and the wider legal profession continue to grapple with how best to integrate rapidly advancing AI tools into teaching and practice. Previous studies have suggested that AI systems can pass bar examinations, achieve top grades in law school assessments, and even assist in grading exams.

An increasing number of law schools are now introducing mandatory AI training in their first-year curricula, although approaches differ. The University of California, Berkeley School of Law, for instance, has recently adopted a policy placing tighter restrictions on how students may use AI in academic work.

The Stanford study suggests AI could play a growing role on the teaching side, offering students on-demand explanations instead of relying solely on peer discussions or occasional faculty email responses.

Importantly, fewer than 4 per cent of AI-generated answers were flagged as potentially harmful to student learning, compared with 12 per cent of answers written by professors.

“We find that, when evaluated by legal educators, AI tutors can offer high-quality, on-demand support that complements classroom instruction and may broaden access to expert guidance,” said study co-author Alejandro Salinas.

For any enquiries or information, contact ask@tlr.ae or call us on +971 52 644 3004. Follow The Law Reporters on WhatsApp Channels.