The rise of AI-powered prototyping tools in UX design has generated excitement, but a recent evaluation shows they often fall short of human capabilities. Researchers at NN/G evaluated these tools by redesigning a profile page for their online training, testing three categories: AI-assisted design tools for static mockups, coding tools for interactive prototypes, and general-purpose chatbots. They used prompts with varying levels of detail, discovering that specificity significantly improved the quality of outputs. For instance, detailed prompts yielded designs that closely resembled human work, while vague prompts led to inconsistent and often off-target results. The study highlighted that AI tools benefit from visual references, such as hand-drawn sketches or Figma links. When these artifacts were included, the AI produced more accurate designs that aligned with the creator's vision. However, even with detailed instructions, AI tools struggled with nuances that human designers typically manage. Common issues included poor visual hierarchy, inconsistent spacing, and color choices that created tension rather than harmony. Although AI can generate high-fidelity designs quickly, these outputs often lack the thoughtful touch that makes a design truly effective. Another critical finding revolves around the inherent bias in AI systems. These tools are trained on vast datasets that favor mainstream design patterns, leading to outputs that tend to be generic and unremarkable. Without specific visual direction, AI-generated designs often default to a minimalistic style, which can come across as flat or interchangeable. The study underscores the importance of understanding AI's limitations—while it can produce designs based on existing patterns, it lacks the intuition and context sensitivity that human designers bring to the table.