AI is increasingly generating creative content, from ads to visual designs, but consumers often respond with aversion. Across six experiments, we show when and why this resistance can be reduced. We reveal a counterintuitive effect of expectancy disconfirmation: AI underperforming increases the likelihood that consumers choose its content (compared with human-generated content), whereas AI overperforming decreases it. This effect is driven by overperformance triggering perceptions of threat and dehumanization, whereas underperformance reduces threat and fosters humanization. These findings challenge traditional views of expectation disconfirmation. We also identify managerially actionable boundary conditions of this effect.
Karpińska-Krakowiak et al. (Thu,) studied this question.