Mapping AI-driven construction risk and safety: a bibliometric analysis (2000–2025)

Abstract

This study maps the evolution of artificial intelligence (AI)-driven risk management and safety research in construction through a bibliometric and science-mapping review of Scopus-indexed journal articles published between 2000 and 2025. From 398 records, screening yielded 272 English-language documents and 177 final journal articles. Performance analysis used Microsoft Excel, OpenRefine, and biblioMagika, while science mapping used the Bibliometrix R package and VOSviewer, supported by manual verification of AI-assisted synthesis. The study integrates performance indicators, network analysis, thematic validation, and gap-oriented interpretation to extend previous bibliometric research on occupational safety and health. Results show a shift from early sensor-based monitoring towards predictive analytics, computer vision, wearable systems, digital twins, conversational AI, and autonomous inspection. Seven clusters were identified: AI safety frameworks; predictive analytics and risk modelling; real-time monitoring and machine safety; worker health monitoring; socio-technical and ethical AI; AI-based safety training and virtual environments; and autonomous inspection. The findings emphasise a transition from reactive practices to proactive, data-driven safety management. Overall, the review provides a structured framework for advancing safer, smarter, and more resilient construction operations.