Basel‑Stadt enables its citizens to participate in development projects that affect their daily lives. However, citizen statements are often abstract or vague, leaving room for interpretation and risking misunderstandings in planning processes. To address this, the thesis investigated whether generative AI tools could reduce ambiguity by converting such statements into images. Additionally, Basel‑Stadt provided discussion protocols from previous redevelopments, which served as a realistic foundation for testing AI image generation in participatory planning.

First, a literature review examined current applications of AI in participatory planning. Second, thematic analysis of Basel‑Stadt discussion protocols identified typical citizen concerns. Both manual coding and AI‑assisted analysis were tested to evaluate automation potential. Third, a comparative analysis assessed four AI tools based on interface, language support, image quality, and editing functions. Finally, practical experiments applied structured, unstructured, and placement‑specific prompts to generate and refine images from citizen statements.

The study shows that AI image generation can enhance participatory processes by making citizen input more concrete and visually accessible. Adobe Firefly emerged as the most suitable tool. However, key limitations were identified. AI‑assisted thematic analysis proved unreliable, requiring manual coding that is accurate but very time‑intensive. Generating images from scratch often led to misplaced or missing elements, and abstract ideas such as “recreational quality” could not be represented convincingly. Despite these challenges, the technology shows promise when used as a complementary tool. For Basel‑Stadt, the main benefit lies in understanding both the opportunities and the current boundaries of AI image generation. The thesis recommends pilot projects in smaller redevelopment contexts, where citizens can directly interact with AI‑generated images and guide iterative refinements. This would eliminate the need for discussion protocols, foster more accurate representations, and provide insights into whether citizens themselves perceive AI‑supported visualisations as useful contributions to participatory urban planning.