Automated compliance checking in AEC in the era of AI and LLMs: A review (2022–2025)

Since 2022, advances in Artificial Intelligence (AI) and Large Language Models (LLMs) have reshaped Automated Compliance Checking (ACC) in AEC. This review applies an AI-driven PRISMA workflow combining LLM-assisted discovery and screening with transparent provenance and audit trails. Studies are mapped to five ACC pipeline stages: rule interpretation, model preparation, rule execution, reporting, and decision support. Iterative coding identifies ten cross-cutting themes used as analytical lenses: Multimodal Information Extraction, Formalisation of Regulatory Text, Semantic Alignment with BIM/IFC, Integration of Ontologies and Knowledge Graphs, Rule Representation and Reasoning, Model-Driven Compliance Intelligence, Tool Development and Real-World Application, Explainability and Trust in AI Systems, Human-in-the-Loop Approaches, Evaluation and Benchmarking. The analysis examines their presence across stages, highlights the rise of LLM-assisted rule discovery, and identifies assurance practices. The review presents a stage-based gap analysis, an evidence-based agenda for interpretable, auditable, multimodal ACC, and a reproducible method for maintaining a living review over time.