Generative artificial intelligence and urban transport resilience: Rethinking sustainable mobility

Urban transport systems face compounding shocks, cascading failures, and rapid context shifts that challenge conventional artificial intelligence (AI) tools designed for structured inputs and stable data regimes. Generative artificial intelligence (GAI) expands disruption management by enabling interaction with heterogeneous information, evidence-grounded synthesis, scenario generation under deep uncertainty, and decision support. This review maps the emerging field and clarifies how GAI can strengthen urban transport resilience while introducing new safety-critical risks. We first conduct a dual-corpus bibliometric analysis of Web of Science Core Collection (2016–2025), covering 1670 AI and resilience publications and a 114-paper GAI subset. We then develop a phase-linked framework that connects four GAI roles — information extraction, knowledge integration, scenario generation, and decision support — to the resilience phases of absorption, adaptation, recovery, and transformation. Synthesising empirical studies across transport operations, planning, and Earth observation (EO) and remote-sensing vision-language model (VLM)–large language model (LLM) pipelines, we find that current evidence is strongest at the capability level, whereas phase-specific transport impacts are less routinely quantified under real operational constraints. Finally, we translate these gaps into an agenda for evaluation and governance, emphasising reliability and uncertainty communication, cybersecurity, data governance and interoperability, and equity-oriented public value.