Abstract: This study reports a bibliometric analysis of a sample of 787 articles from the Web of Science database. Railway safety is a broad research field with high research activity. Computer-assisted techniques such as bibliometric analysis could be used to study the thematic evolution and intellectual structure of such high-activity research fields. This study aims to identify recent advances in railway safety research and suggest future research directions. We first set out to identify the themes that define railway safety research through science mapping. Using citation count and PageRank of articles, the most prestigious articles published within the last decade were identified for content analysis. Studies on artificial intelligence applications to railway safety, particularly in hazard identification, infrastructure and equipment control, and condition monitoring, have gained traction in recent years. The security and safety of proposed cyber-physical systems for railway applications have been widely studied. Studies on next-level train control systems such as ETCS-3 and CTCS-4 have focused on satellite-based train localization and 5G technology (and beyond) for railway communication systems. Research on human reliability in railway operations and management has also received considerable attention from researchers. However, little consideration is given to addressing issues related to the reliability of expert judgments–a key instrument in practical railway safety assessment. In the end, we suggest six avenues for future research in railway safety.