Can We Tell the Difference? A Turing Test on Human Perceptions of Innovation Ideas in Text Created by ChatGPT

ABSTRACT Generative artificial intelligence (GenAI), such as ChatGPT, is increasingly recognized for its potential to drive innovation, yet questions remain about its ability to mimic human innovation. Leveraging cognitive load theory (CLT), this study explores whether humans can reliably distinguish between incremental and radical innovation ideas in text created by ChatGPT and those created by humans. In an online experiment with 689 participants and 6454 innovation ratings, results from Turing and noninferiority tests showed that participants achieved accuracy only slightly above chance level, with marginally higher accuracy for radical innovations. However, significant differences emerged across thematic clusters: Convenience and health innovations were identified more accurately than education and sustainability innovations. This research advances the understanding of how closely GenAI‐generated innovations resemble human ones, emphasizes the need for transparency in GenAI's role within creativity processes and raises questions about developing new indicators for evaluating contributions in increasingly AI‐integrated innovation environments.