Large language models in building energy applications: a survey
Muhammad Arslan, Saba Munawar
Abstract
The use of large language models (LLMs) in building energy applications (BEAs) is driving intelligent and sustainable solutions. Research in this area has expanded across multiple subfields, highlighting the need for a comprehensive understanding of LLM adoption, key applications, and emerging trends. Existing surveys often focus on narrow technical use cases, overlooking the broader context of LLM integration in building energy (BE) systems. This survey reviews 76 peer-reviewed articles published between 2021 and July 2025 at the intersection of LLMs and BEAs. A multi-scale analysis is presented, including keyword analysis, conceptual linkages via co-occurrence networks, topic modelling across six domains, and temporal assessment of study and method distributions. This approach provides a structured synthesis without delving into model-specific technical details. Key findings indicate that LLMs are transitioning from experimental tools to core infrastructure: they serve as semantic connectors between modelling, automation, and human-centred feedback; foundational methods dominate topic share; and methodological maturity has accelerated since 2023. Practical applications include semantic data integration, automated occupant surveys, decision support for retrofits, and energy-aware control. The survey offers a roadmap for scalable, interoperable, and human-aware BEAs, informing both research and practice.