This study evaluates the capability of Generative AI (GenAI) tools in determining urban perception through street-level imagery (SLI) in the central part of Ankara. Specifically, Gemini 1.5 Flash and Large Language Vision Assistant (LLaVA) 13B models are utilized to assess predefined perception aspects. While Gemini operates via an application programming interface (API), LLaVA is executed locally using the Ollama platform. The findings indicate that Gemini 1.5 Flash provides consistent results across English, German, and Turkish prompts, whereas LLaVA 13B struggles with non-English queries and occasionally returns NA values for specific perception aspects. Beyond model performance, this study explores the potential contributions of GenAI to urban perception mapping. By integrating auxiliary data such as air quality, noise levels, and temperature, GenAI could enhance urban analytics beyond visual interpretation. However, challenges remain, including limitations in capturing personal experiences, cultural backgrounds, and subjective perception variations. Despite these constraints, GenAI presents a promising approach for large-scale, automated urban perception assessment, supporting sustainable and inclusive urban planning. Future studies should focus on refining multimodal GenAI models, incorporating diverse data sources, and addressing biases to improve the accuracy and applicability of urban perception analysis.