The utilization of generative artificial intelligence (AI) in architectural design has established a new paradigm in the relationship between the designer’s mind and the visual form. While current research predominantly focuses on optimization, AI's potential to decode theoretical discourses and transform abstract concepts into spatial representations remains understudied. This study investigates the correlation between textual perception and AI-generated images, using the complex, metaphorical spaces in Jorge Luis Borges's literature as a case study to challenge the limitations of visualization.
Adopting a qualitative approach based on content analysis and visual representation, spatial descriptions were systematically extracted from six Borges short stories and inputted as prompts into Google Gemini. The generated images and primary texts were then analyzed and coded using ATLAS.ti software. To ensure validity, a triangulation strategy evaluated AI visual fidelity across six distinct image-generation scenarios.
The findings demonstrate that AI successfully transformed textual codes into visual equivalents, recreating the text's "architectural essence" within standard classifications like geometry, materials, and qualitative attributes. Consequently, rather than being a mere translation tool, AI functions as an intellectual assistant capable of comprehending the text's mental structures and extending them into unwritten spaces.