Leveraging LLMs for Efficiency in Family Law Research

This paper investigates the use of Large Language Models (“LLMs”) to support legal research workflows in the domain of Singapore family law, specifically in divorce-related matters involving the division of matrimonial assets and child custody. In common law jurisdictions, legal research is particularly labour-intensive, as lawyers must extract legal principles from complex and often lengthy judicial decisions. This reliance on precedent makes case law a natural candidate for structured automation. However, the legal sector has been slow to adopt AI due to concerns over reliability, explainability, and professional trust.

This paper presents an automated research pipeline that integrates web scraping, LLM-based judgment analysis, and structured data extraction, supported by a verification mechanism to flag high-risk outputs and link them back to source text. The system achieved an average extraction accuracy of 96.67%, outperforming a human benchmark of 84%, while maintaining transparency through a “human-in-the-loop” review framework. The findings suggest that LLMs, when applied responsibly, are well-suited for upstream legal tasks such as case triage, intake, and research preparation—providing a viable and scalable approach to improving access to justice through the augmentation of legal research in common law systems.