Advancements in artificial intelligence (AI) present numerous opportunities to routinize and make the law more accessible to self-represented litigants, notably through AI chatbots employing natural language processing for conversational interactions. These chatbots exhibit legal reasoning abilities without explicit training on legal-specific datasets. However, they face challenges processing less common and more specific knowledge from their training data. Additionally, once trained, their static status makes them susceptible to knowledge obsolescence over time. This article explores the application of retrieval-augmented generation (RAG) to enhance chatbot accuracy, drawing insights from a real-world implementation developed for a court system to support self-help litigants.