Blind study finds AI prototypes usable but rated less original

Karoline Romero, Igor Wiese, Renato Balancieiri, Gislaine Camila Leal and Guilherme Guerino, presenting at SEMISH 2026, ran a blind study in which 92 participants rated AI-generated and human-made interface prototypes without knowing which was which, scoring them on the UEQ-S scale for both pragmatic and hedonic quality. The pragmatic dimensions, usability and efficiency, came back positive for the AI output; the hedonic ones, originality and innovation, came back neutral or negative. That split matters more than a single average score would: it suggests GenAI tools can produce something that works but defaults to familiar layouts and conventions, which participants read as unoriginal even when they rated it functional. For teams treating AI prototypes as a starting point rather than a finished design, that is closer to what to expect than the more sweeping claims usually made for these tools.