Generative artificial intelligence in architectural education: An experimental study on the early design process

In this thesis, it is aimed to determine the advantages and difficulties that may be encountered in integration of artificial intelligence (AI) technology, which is rapidly increasing in use today, in architectural education. Within this framework, the research was carried out in a three-stage process, focusing on the use of generative AI technology in the preliminary design process.In the first stage, a literature review was carried out to establish the theoretical framework, followed by a bibliometric analysis to identify current themes and research trends. The results of the systematic analysis revealed that generative design has emerged as a current and prominent topic of discussion in the use of AI within architectural practice. In the final stage, an experimental study was conducted in which students used Autodesk Forma (Site Automation module), Archistar, and OneClick LCA software extensions to experience generative AI tools in the early design phase. Data collection included pre-test and post-test surveys, student feedback, and process observations. The scales based on the Technology Acceptance Model (TAM) were tested for reliability using Cronbach's Alpha, and the data were further analyzed through statistical methods (descriptive statistics, Chi-square, and Friedman tests) as well as qualitative evaluations. The findings indicate that AI in architecture is used for generative design, optimization, performance analysis, alternative scenario generation, environmental data analysis, and sustainability assessment, while in architectural education it supports conceptual thinking, creativity, visualization, and data-driven decision-making. In the early design phase, AI tools contributed significantly to mass placement, geometry decisions, environmental analyses, density, and functional distribution. On the one hand, students emphasized advantages such as accelerated idea generation, the ability to produce alternatives, time efficiency, parametric flexibility, and scenario diversity. On the other hand, challenges were identified, including usability difficulties, interface limitations, lack of local databases, contextual mismatches, and concerns about AI constraining creativity. The unique contributions of this study include emphasizing the necessity and increasing the use of different generative artificial intelligence tools in the design process, evaluating student experiences, and using local contextual data (topography, existing building density, green space information, etc.) in Türkiye. The limitations of the study are related to the small sample size and its exclusive focus on the early design mass production phase. For future research, it is recommended to explore other phases such as conceptual, planning, and facade design, to develop local data infrastructures, and to strengthen interdisciplinary and pedagogical collaborations. KEYWORDS: Architectural Education, Artificial Intelligence, Generative Design, Experimental Study, Form Finding