Integrating Artificial Intelligence and BIM in Construction: Systematic Review and Quantitative Comparative Analysis
Reinaldo Valdebenito, Eric Forcael
Abstract
In the transition toward a more digital and data-driven construction industry, understanding how Artificial Intelligence (AI) and Building Information Modeling (BIM) are integrated is key to planning, delivering, and operating projects effectively. This review examines recent studies to identify usage patterns of AI and BIM. Searches were conducted on the Web of Science Core Collection from 2022 to 2025. After running a reproducible review protocol aligned with PRISMA 2020, which began with 1212 articles, and after a funneling process, 12 studies met the predefined eligibility criteria. In the present study, the synthesis was non-meta-analytic; instead, the information was analyzed by using standardized tabulation with a consistent format and compared using a two-level weighting scheme. The methodological approach combines full-text reading and descriptive coding with a reproducible weighting scheme that accounts for mentions per study and integrates them at the corpus level using open-source tools. The results show a strong focus on Deep Learning (DL), with a greater presence in Digital Twins (DT) and BIM Modeling (BIMM); Multidimensional BIM (4D/5D) appears as a secondary line, while the Common Data Environment (CDE) and Clash detection (CD) are sporadic. The coupling of DL-DT and DL-BIMM predominates. Simultaneously, Machine Learning (ML) provides explainable analysis on structured data, and Generative Adversarial Networks (GAN) and Automated Machine Learning (AutoML) with Machine Learning Operations (MLOps) act as enablers for data generation/adaptation and deployment with traceability. It is concluded that advancing metrics and shared datasets, especially for CDE and CD, along with developing reproducible workflows oriented toward MLOps, are key to scaling AI in real-world environments.