Problem statement: The widespread use of chatbots in hospitality and tourism leads to inevitable service failures. Although research has investigated the influence of chatbots` anthropomorphic cues, comprehending how distinct anthropomorphic cues influence user behavior in service failure is still limited. Methodology: To explore how the anthropomorphic design of chatbots affect user`s tolerance for service failure, this research conducts a 2 (anthropomorphic appearance: 3D vs. 2D) x 2 (language style: informal vs. formal) x 2 (interdependent self-construal: high vs. low) between-subject online experiment. Result: Results show that the congruent anthropomorphic cues of chatbots can significantly improve consumers ' tolerance, where attachment mediates this process positively. Additionally, the interdependent self-construal level plays a positive moderating role in this process. Implications: This study contributes theoretically by explicating anthropomorphism in attachment and cognitive dissonance theory and extending the understanding of self-construal theory. Moreover, the study provides recommendations for managers to design effective anthropopathic chatbots.