Large Language Models in Legal Analysis

In 2020, we published a comprehensive overview of the landscape of computational legal analysis. Since then, the emergence of large language models (LLMs) has fundamentally transformed the field. While traditional computational methods in law focused on relatively simple text processing and statistical techniques, LLMs now enable more advanced operations on legal texts that traditionally required expert legal reasoning--often approaching or even surpassing human-level performance on these tasks. This article provides an updated framework for understanding how LLMs are reshaping legal scholarship and practice. We examine three primary applications of LLMs in legal analysis: automated feature extraction and data generation from legal documents; prediction, classification, and description tasks; and engineering-focused studies that develop benchmarks and optimize model performance for legal domains. The rapid advancement of LLMs presents both unprecedented opportunities for scaling legal analysis and significant challenges around accuracy, bias, and professional responsibility. As these models become increasingly integrated into legal practice, understanding their capabilities and limitations has become essential for legal scholars, practitioners, and policymakers.