AI is officially better at legal research than lawyers.

A new Vals Legal AI Report has found that four AI tools, like Counsel Stack, Alexi, Midpage and ChatGPT, outperformed a lawyer control group across 200 legal research questions.

All AI tools topped the human baseline of 69%, with the legal-specific AIs scoring 76–78%, and ChatGPT close behind at 74%. Accuracy was high — roughly 80% for legal and general AIs, compared with 71% for human lawyers.

But the key differentiator was citations.

Legal-specific systems scored six points higher on authoritativeness, drawing on proprietary databases — even if much of the data is publicly available.

The study, designed with input from Paul Weiss, McDermott, Reed Smith and Paul Hastings, found that AI outperformed lawyers on three-quarters of all questions, sometimes by margins exceeding 30 points.

But luckily, humans still held an edge where context, nuance and multi-jurisdictional judgment were required. We’ll take that as a win.

Several major legal AI companies declined to participate, despite appearing in Vals’ earlier report on tasks like redlining and document review. The study also relied on “zero-shot prompts” — single, context-free questions — which Vals concedes don’t reflect how lawyers or AI tools typically handle real research workflows.

Still, it’s a hard pill to swallow.

AI isn’t just matching lawyers. It’s overtaking them.

For Australian firms, that’s a wake-up call. Those investing early in AI-ready infrastructure and platforms like Harvey, Legora and CoCounsel will likely seize the lead, while others risk being left behind in law’s next productivity boom.

Source: Vals Legal AI Report