As ecological principles increasingly guide contemporary architectural and urban design, optimizing green-space layouts has become central to improving built-environment performance. This study applies the Pix2Pix model to interpret planning-area maps and develops a green-space layout support system implemented in a Unity3D engine with a Python-based workflow. The system enables visual simulation of alternative green-layout schemes and helps designers compare options to identify a preferred configuration. In a case validation, the AI-assisted optimization adopts a zoned daylighting strategy for public areas. Simulation results show reductions in district-building energy use, with cooling demand decreasing from 75.64 kWh/m² to 65.32 kWh/m² and heating demand dropping from 45.26 kWh/m² to 42.31 kWh/m². Residents also reported high satisfaction with the simulated retrofit outcomes. Overall, the proposed approach supports environmentally responsible and sustainable smart-city development by lowering environmental impacts and contributing to healthier, more livable urban spaces.