While Generative Artificial Intelligence (GenAI) has shown great advancement in various aspects such as text, image, and video generation, its application in user interface (UI) design remains relatively underexplored. This study examines the capabilities of GenAI tools to produce UIs that meet established design standards. An experimental framework was employed, consisting of six structured tests across four GenAI systems — ChatGPT, Microsoft CoPilot, Google Gemini, and Google Stitch — covering image-based and markup-based generations of UIs. The outputs were evaluated using criteria including visual fidelity, structure, markup code quality, and trueness to the prompts. The findings show a divide between visual ideation and functional implementation. Although generalist GenAI models can generate aesthetic visual mockups, they fail to produce production-quality markup code. Only the specialised tool, Google Stitch, demonstrated the capacity to consistently generate high-quality UIs across most tests. This suggests that reliable and consistent UI generations are currently better achievable through specialised systems. The research concludes that GenAI can generate UIs. However, they serve more effectively as a creative assistant for designers than a replacement for developers as far as UI design implementation goes.

Keywords: Generative AI, User Interface Design, AI-Assisted Design, Human-Computer Interaction.