Large-scale analysis of environmental and circularity metrics in construction products through automated data extraction using large language models

The selection of construction products significantly impacts building practices as well as environmental and biodiversity outcomes. However, current Environmental Product Declarations (EPDs) lack standardized circularity metrics and contain highly heterogeneous, manufacturer-dependent data, which severely limits sector-wide assessments and data-driven decision-making. Existing methods fail to systematically extract and structure this fragmented information at scale, creating a critical barrier to benchmarking circularity-related metrics, such as circular origin, dismantlability potential, and end-of-life pathways, and environmental performance across construction products. This study addresses this gap by introducing a novel Large Language Model (LLM)-based approach for the reliable extraction and standardization of data from diverse EPDs. Applying this method to 8463 products reveals three key findings: (i) manufacturers deprioritize circular end-of-life pathways; (ii) environmental and biodiversity impacts vary substantially both within and across product categories based on material composition; and (iii) a “circularity paradox” persists, whereby products sourced circularly often end up in landfills. Methodologically, this research establishes the first scalable procedure for comprehensive benchmarking of circularity-related metrics and environmental performance in construction products. Practically, it identifies critical intervention points including the standardization of circularity and environmental reporting in EPDs, the promotion of digital product passports, and the reinforcement of industrial symbiosis with transformation companies. For policymakers, the results highlight how existing regulatory shortcomings perpetuate linear economic models and obstruct consistent, equitable reporting, highlighting the urgent need for more stringent, standardized disclosure requirements to enable nuanced environmental comparisons. • Novel LLM–EPD approach for mapping end-of-life circularity in construction sector. • Manufacturing phase (A1–A3) dominates both environmental and biodiversity impacts. • Some products show circular paradox: recycled origin but landfilled at end-of-life. • Large environmental variability found within product categories due to materials. • More robust EPD validation and standardization needed for reliable comparisons.