[X]score: 0.51
Gemini 2.5 Pro Beats Law Professors 75% of Time in Blind Study
June 2, 2026
A Stanford study had law professors blind-evaluate office-hours answers from peers versus Gemini 2.5 Pro, with Gemini winning 75% of comparisons and rated less harmful than human responses. Newer models performed even better on the same tasks.
HOW THIS AFFECTS YOU
●
researcherBlind evaluation methodology and 75% win rate provide a concrete benchmark for LLM performance on expert legal Q&A worth replicating in other domains.
●
founderWorth watching because legal AI products can now point to a Stanford-backed blind study showing Gemini 2.5 Pro outperforms human experts, lowering sales friction.
●
policyGemini answers rated less harmful than human professors complicates simple narratives about AI safety risk in high-stakes advisory contexts.