When Chatbots Cause Trouble: How Agent Type and Conversation Style Shape Consumer Responses to Service Failures

Purpose The primary objective of this study is to investigate the consumer response to service failure involving different types of service agents – specifically human agents, anthropomorphic chatbots, and non-anthropomorphic chatbots- in the services sector. In addition, the research underscores conversation style and service criticality as key moderating factors influencing consumer responses to chatbot service failure. Design/methodology/approach. The hypothesised relationships are tested across three between-subjects experiments. Study 1 (n=296) and study 2 (n=169) were conducted in the context of restaurant table reservations, while study 3 (n=176) involved booking doctors’ appointments in a healthcare setting. Findings Customers showed stronger brand avoidance when they encountered a service failure by a virtual agent than by a human agent. This effect was stronger for a service failure by a non- anthropomorphic chatbot than by an anthropomorphic chatbot. Brand avoidance was exacerbated for high-critical services. Moreover, customers using anthropomorphic chatbots with a social message orientation demonstrated lower brand avoidance than with anthropomorphic chatbots with a task-oriented message. Originality This research provides new insights by revealing that service criticality and conversation style act as key boundary conditions in the link between chatbot failure and brand avoidance — an area previously unexplored. Moreover, by drawing on stress-coping theory, this research suggests that psychological distress is s a novel mediating mechanism, demonstrating its pivotal role in shaping consumers' brand avoidance following chatbot service failures. This innovative perspective deepens our understanding of chatbot-related service failures in multiple contexts and offers key insights into the mitigation of brand avoidance. Practical implications Managers should prioritize the adoption of anthropomorphic chatbots over non- anthropomorphic ones to mitigate brand avoidance, as customers are more likely to avoid brands when less human-like chatbots are responsible for the failure. Additionally, firms should carefully design chatbot interactions in service failure contexts, high in criticality, where reliance on automated agents may heighten consumer discomfort and lead to brand avoidance. Social implications Our findings suggest that service failures by chatbots cause psychological distress in consumers, suggesting that caution is needed when considering the complete replacement of human service agents with chatbots.