Artificial intelligence in litigation: a systematic review of applications, challenges, and future directions

Abstract The integration of artificial intelligence (AI) into litigation has accelerated in recent years, driven by advances in natural language processing, deep learning, and large language models. While AI promises to enhance judicial efficiency, decision consistency, and access to justice, its deployment in litigation raises substantial technical, institutional, and trust-related challenges. This paper presents a systematic review of AI applications in litigation, analyzing 90 peer-reviewed studies published from January 2020 to 3rd of December 2025 and following the PRISMA 2020 guidelines. We categorize existing legal AI systems into five key application domains, including AI-based online dispute resolution systems, legal information and knowledge retrieval systems, document analysis and text-mining systems, expert and consultation systems, and judgment and outcome prediction systems. The review synthesizes dominant model architectures, datasets, and evaluation practices, highlighting a clear transition from rule-based approaches to transformer-based and large language models. Despite these advances, we identify persistent challenges related to data scarcity, limited generalizability, explainability, regulatory uncertainty, cybersecurity, and the risk of hallucination in legal AI systems. Based on the evidence reviewed, we argue that future progress in AI-assisted litigation depends on the development of domain-specific, multimodal models; robust explainability mechanisms; privacy-preserving learning frameworks; and clearer regulatory guidance. This review provides a structured reference for researchers, legal practitioners, and policymakers seeking to design, evaluate, and govern trustworthy AI systems for litigation. The limitations of the review are mainly about the search bias in the search, retrieval, and selection strategy, and potential bias in database and language selection.