Large language models as enablers of industrial information integration in construction: an adoption mechanism study

Large language models (LLMs) are emerging enablers of industrial information integration, yet their adoption mechanisms in construction remain unclear. This study develops an integrated framework that combines the Technology Acceptance Model and the Information Systems Success model, extended with trust, prior LLM experience, perceived substitution crisis, and job specialization, to explain individual adoption of LLMs for information integration in construction. Data from a cross-sectional survey of 82 Chinese construction professionals are analyzed using partial least squares structural equation modeling and bootstrapping. Results show that perceived ease of use robustly strengthens attitude and provides directional evidence for strengthening perceived usefulness, while perceived usefulness strengthens attitude and attitude drives intention to use. Trust, reinforced by perceived information quality, shows a positive association with intention under the primary one-tailed criterion, but this relationship requires further validation under the two-tailed robustness check. System quality and information quality indirectly foster adoption by improving perceived ease of use, although the relationship between system quality and trust is supported only as directional evidence, while prior LLM experience enhances ease of use but not perceived usefulness or trust. Perceived substitution crisis and job specialization exhibit no significant effects. The findings highlight that successful LLM-enabled information integration depends on both technical quality and psychological appraisals, and they offer implications for designing trustworthy, humancentered LLM applications that advance intelligent, interconnected construction ecosystems. The study links micro-level user acceptance with macro-level digital transformation and integration strategic objectives.