Satisficing Justice: Generative AI as Legal Triage

This article uses bounded rationality and satisficing to reassess the value of generative AI legal consultation for access to justice. Based on 480 tests across three representative civil and criminal scenarios and field notes from public legal service agencies and courts in two cities in Southwest China, it argues that generative AI both enables and constrains grassroots legal services. DeepSeek-V3.1 performs stably on core legal-liability judgments and can provide cognitively constrained users with timely preliminary information that satisfies early-stage decision needs, yet fieldwork shows that self-service terminals remain difficult for many users because of typing barriers, weak problem formulation, dialectal speech recognition, and limited interaction support. In ordinary Chinese cities where lawyer quality is uneven and some low-end services are misleading or fraudulent, AI should not be understood as a substitute for high-quality lawyers; its practical value lies in providing a relatively neutral baseline that can function as a substitute for poor legal services and as a defensive tool before users enter risky representation relationships.