A recent study unveiled by Vals AI has sparked interest across the legal tech community by suggesting that both specialized legal and general artificial intelligence (AI) models now potentially rival, or surpass, human lawyers in the realm of legal research accuracy. The report, known as VLAIR – Legal Research, builds on the groundwork laid in Vals AI’s previous report from February 2025.
This study critically assessed the capabilities of three legal AI systems—Alexi, Counsel Stack, and Midpage—and one general AI model, ChatGPT, by juxtaposing them with a baseline of traditional lawyer-led research. The findings reveal that all AI systems, including ChatGPT, performed at an accuracy level comparable to human researchers, with the legal AI solutions slightly edging out in overall performance.
More intriguing is the study’s uncovering of scenarios where the AI models demonstrated superior swiftness and accuracy, notably in tasks demanding the summarization of legal holdings and sourcing recent case law. This was evident as AI models completed tasks in a matter of minutes compared to the more prolonged response times of their human counterparts.
Nonetheless, despite the performance of these AI solutions, a significant talking point is the non-participation of some leading AI research platforms like Thomson Reuters and LexisNexis. vLex initially signed up, but later withdrew, attributing its withdrawal to discrepancies in study design, which it argued was not adequately tailored for enterprise AI tools.
The Vals AI benchmark measures AI capabilities across several weighted criteria—accuracy, authoritativeness, and appropriateness. This comprehensive evaluation highlighted that while both legal and general AI models manage to provide accurate answers, the legal-specific tools maintain an advantage in authoritativeness due, in part, to access to proprietary legal databases.
Yet, the report also points out AI’s limitations in handling complex, multi-jurisdictional questions, areas where human expertise continues to prove invaluable. These findings underline the important balance between the growing integration of AI in legal processes and the enduring need for human judgment in complicated legal reasoning.
Although the absence of data from the major contributors leaves gaps in the findings, the VLAIR – Legal Research study nonetheless provides a critical perspective on the evolving role of AI in legal research, serving as a benchmark for understanding the potential and limitations of AI as it stands today within the legal domain.