Observations visualisation tendencies with Text-to-image GenAI

Text-to-image GenAI continues to draw interest among designers for translating ordinary text into sophisticated visuals instantly. As such, previous research had explored the potential application of Text-to-image GenAI across different design processes and project phases. More recent research have begun to demonstrate the implications of Text-to-image GenAI on current and future design interactions. However, research alike has yet to compare the implications of Text-to-image GenAI on individuals from diverse disciplinary backgrounds. This reveals a knowledge gap in design research as the democratisation of visualisation is regarded as one of the most significant characteristics of Text-to-image GenAI tools used in design today. The research presented in this paper responds by comparing architect and non-design engagements with Text-to-image GenAI to understand the implications of Text-to-image GenAI on the early establishment and management of client-architect relationships. To do so, design experiments were conducted where individual architects and non-designers received basic training and were given the task to visualise architectural design ideas using Midjourney. Findings from this study revealed visualisation tendencies of concept reframing, additive ideating and divergent-convergent preference developing from observed design idea prompting and selecting strategies. However, no particular visualisation tendency were observed to dominate in one participant group over the other. This suggests that architects and non-designers share similar approaches to visualising ideas through Text-to-image GenAI. As such, uncovering future opportunities to investigate the process of visualisation that were previously seen to limit the inclusion of clients and other stakeholders in the establishment and management stages of client-architect relationships in architecture.

keywords: generative artificial intelligence; design visualisation; text-to-image genai; architecture

• with: Nadia Anam (lead), Charlie Ranscombe
• year: August 2026 — ongoing