Why AI design tools keep making the same flyer (grafista.io)

A recent test has revealed that generative AI design tools, namely Claude Design, Canva, and Moda, often produce repetitive flyer designs when given the same prompts, limiting their creative output to variations of a singular house style. In contrast, Grafista differentiated itself by generating six unique designs from the same prompts, showcasing a diverse range in color palettes, layouts, and typographic choices. This discrepancy is attributed to the way these models optimize for output based on what they have encountered most frequently in their training data—a behavior known as "mode collapse"—which leads to a convergence on familiar design patterns unless explicitly directed otherwise. This finding is significant for the AI/ML community as it highlights the limitations of current AI systems in creative fields, particularly in graphic design. The test emphasized the importance of specific prompts to elicit varied outputs, suggesting that users need substantial design knowledge to navigate these tools effectively. Grafista's approach, which relies on visual Blueprints rather than textual prompts, allows for a more nuanced and flexible output, providing a potential solution for users seeking creativity without in-depth design expertise. The study serves as a critical examination of how generative AI can adhere to defaults and how alternative methodologies can enhance creative expression in design applications.