Artificial intelligence for decoding the cultural “genes” of traditional architecture: a systematic review of façade, spatial configuration, style, and semantics

Cultural “genes” embody the material expressions and spiritual essence of regional architectural culture. Yet conventional manual decoding remains inefficient, subjective, and weakly integrated across multiple gene dimensions. This study adopts a hybrid review framework (systematic literature review, CiteSpace-based bibliometric visualization, and coding-matrix synthesis). Based on 48 core publications (2015–2025) from Web of Science and Scopus, it maps the evolution of AI-enabled cultural-gene decoding in traditional architecture, identifies major gaps, and clarifies the field’s paradigm shift. Results show a steady increase in publications. Eastern studies emphasize four-dimensional integration, whereas Western research focuses on style-oriented decoding. Methodologically, computer vision and deep learning dominate, increasingly combined with 3D point-cloud processing and semantic segmentation. The field is shifting from experience-based qualitative description and typological classification toward data- and algorithm-driven association mining, quantitative explanation, and generative translation, but challenges remain in cross-dimensional coupling, interpretability, and application generalization. To address these issues, this review proposes a four-layer collaborative framework (Input–Technology–Decoding–Output) and a maturity-scale instrument, empirically validated through case studies. It also summarizes four practical pathways for conservation, design translation, digital transmission, and interdisciplinary education.