Large Language Models Are Learning to Reason About Buildings, Review Finds

Large language models are moving beyond text processing into active cognitive partners capable of spatial reasoning across the building lifecycle, according to a systematic review published in Artificial Intelligence Review.

The paper, authored by Gao Xinting, Wang Yating, Miao Zhijian and Zhuang Weimin, analyzes 76 representative studies examining how large language models enable spatial intelligence across the architecture, engineering and construction industry.

The review proposes a layered spatial intelligence framework spanning design, construction, operation, and research and education, offering a structured way to organize how these models contribute at each stage of a building’s life.

According to the authors, the field is undergoing a digital transformation driven by the need to integrate fragmented, heterogeneous data into actionable intelligence, a challenge that large language models are increasingly positioned to address.

Key findings highlight the effectiveness of these models in requirement translation, multimodal design generation, regulatory compliance, and lifecycle Building Information Modeling integration.

Despite these advancements, the review identifies significant challenges that persist, including spatial reasoning accuracy, the lack of physical embodiment, and institutional data governance.

The authors outline a strategic roadmap toward embodied intelligence, AEC-specific world models, and enhanced human-AI collaboration, positioning the framework as a foundation for advancing LLM-driven spatial intelligence in professional practice.

The article is published open access, with the version shared early as peer-reviewed, accepted research carrying a permanent DOI and subject to further edits before the final Version of Record.

Subject of Research: A systematic review of how large language models enable spatial intelligence across the architecture, engineering and construction lifecycle.

Article Title: Enabling spatial intelligence in the AEC lifecycle: a framework and review of large language models

Article References: Xinting, G., Yating, W., Zhijian, M., & Weimin, Z. (2026). Enabling spatial intelligence in the AEC lifecycle: a framework and review of large language models. Artificial Intelligence Review. https://doi.org/10.1007/s10462-026-11737-w

Image Credits: AI Generated

DOI: 10.1007/s10462-026-11737-w

Keywords: large language models, spatial intelligence, AEC industry, BIM, systematic review, architecture, construction, design generation, regulatory compliance, human-AI collaboration, embodied intelligence, Artificial Intelligence Review