Service type conditions stance attribution effects on switching after generative AI failures

Generative artificial intelligence (GenAI) service failures can prompt users to abandon a provider. Drawing on the Stereotype Content Model, with the Associative-Propositional Evaluation Model as a complementary process perspective, this research examines how intentional and design stance cues shape warmth, competence, and switching intention across service types. Study 1 used an ERP experiment and found stance-cue x service-type interactions for P2 and P3. Holm-adjusted follow-up tests showed that no P2 simple contrast remained significant, whereas mechanical failures elicited a larger P3 than emotional failures under design stance cues. Study 2 used a scenario experiment and found that intentional stance cues increased perceived warmth and competence, but only competence mediated lower switching intention. This conditional indirect effect was significant for mechanical tasks but not emotional tasks, and the warmth pathway was not significant. Together, the findings identify competence as a task-contingent pathway linking stance cues to switching intention and show that the effectiveness of stance framing depends on the service context.