From the perspective of Dual-Process Theory, this study integrates Social Presence Theory and Arousal-Biased Competition Theory, and adopts a combination of scenario-based experiments and event-related potential (ERP) experiments to examine the effect of anthropomorphism level of tourism chatbots on users’ forgiveness intention and its underlying mechanism. The results show that highly anthropomorphic chatbots improve users’ forgiveness intention by enhancing perceived competence and social presence. The interaction effect between anthropomorphism level and service type on perceived competence is significant. In hedonic services, high anthropomorphism significantly improves perceived competence, while this effect is not significant in utilitarian services. The mediating effect of anthropomorphism level on forgiveness intention through social presence is not moderated by service type. ERP results show that attention resource allocation is significantly higher under the low anthropomorphism condition, cognitive conflict is significantly higher under the high anthropomorphism condition, and heuristic neural responses act as cognitive anchors and bias the subsequent systematic evaluation process.