AI and the syntax of tradition: quantifying morphological fidelity in generative representations of vernacular housing

OpenAlex · İngilizce

This study investigates the capacity of text-to-image (T2I) artificial intelligence models to reproduce the morphological characteristics of vernacular architecture across culturally distinct contexts. Focusing on Turkey, Japan, and Mexico, a comparative framework is developed based on a standardized prompt system and a visual morphological analysis matrix. The study integrates expert evaluation with the Analytic Hierarchy Process (AHP) to assess form typology, material articulation, roof configuration, spatial organization, openings, and symmetry. Results reveal that while contemporary T2I models can successfully approximate dominant formal and material cues, they often simplify deeper spatial and cultural logics embedded in vernacular traditions. Among the evaluated platforms, Leonardo.Ai demonstrates the highest level of morphological fidelity, followed by DALL·E 3, while other models show increasing reliance on stylistic generalization. The findings position generative AI as a selective mediator of architectural knowledge and propose a replicable framework for evaluating cultural fidelity in AI-generated design outputs.