EVALUATING GENERATIVE AI MODELS FOR HISTORIC URBAN HERITAGE VISUALIZATION IN SMART CITIES

Abstract

Evaluating generative AI models across five sites, this study investigates historical urban heritage visualization on Cairo’s Al-Mu'izz Street to advance smart city preservation frameworks. Findings demonstrate that Gemini 3.1 Pro achieves superior conservation fidelity (84.9% overall; 4.20/5 in retention and historical reliability), whereas ChatGPT 5.6 Sol excels in visual quality and material expression. Cross-model optimization reveals that ChatGPT-derived instructions paired with Gemini execution yield the highest authenticity (89.8%). Ultimately, integrating structured, AI-assisted prompt workflows into digital twin environments provides urban planners with a scalable, historically reliable mechanism for heritage conservation and urban management.