A Systematic Review of Large Language Models for Automation in Civil Engineering: Applications, Challenges, and Future Directions

The swift progress of large language models (LLMs) has generated substantial interest in their capacity to revolutionize automation in diverse fields, such as civil engineering. Although large language models have shown impressive abilities in processing natural language and automating tasks, their potential in civil engineering has not been thoroughly investigated, with research remaining scattered and lacking comprehensive consolidation. This paper presents a thorough systematic review to chart the existing scope of LLM-based automation in civil engineering, with the aim of uncovering primary applications, obstacles, and prospective research avenues. We analyze existing studies across multiple dimensions, such as civil and structural engineering, industrial automation, traffic management, education, scientific research, and software development, then critically evaluate the methodological approaches and practical implementations reported in the literature. The review indicates LLMs hold potential for automating design optimization, construction planning, and decision-making processes, but struggle with issues such as gaps in domain-specific knowledge, poor data quality, and safety risks. Moreover, we pinpoint developing tendencies, such as the merging of LLMs with digital twins and building information modeling (BIM), which may transform automation in the domain. The findings highlight the need for robust evaluation frameworks and interdisciplinary collaboration to address technical and ethical barriers. This review consolidates these insights, establishing a basis for subsequent investigations and the actual implementation of LLMs in civil engineering automation.