The dark side of AI customer service: How service failures damage consumer-brand relationships

Abstract

Artificial intelligence (AI) customer service is increasingly used as a frontline interface, yet its failures may create relational damage that extends beyond a single service encounter. Drawing on service failure theory and consumer-brand relationship literature, this research examines how AI service failures are formed and how they influence negative consumer-brand relationships. Study 1 collected 3208 online complaint records from a major Chinese consumer complaint platform. After data cleaning, 2824 valid complaints were analyzed using Latent Dirichlet Allocation topic modeling, which identified ten complaint topics and further synthesized them into five AI service failure dimensions: functional failure, informational failure, systemic failure, interactional failure, and service recovery failure. Study 2 used survey data from 965 Chinese consumers and applied PLS-SEM, Artificial Neural Network (ANN), fsQCA, and Necessary Condition Analysis (NCA). The results show that all five failure dimensions positively influence negative consumer-brand relationships. Informational failure and systemic failure emerge as particularly influential across linear, predictive, and configurational analyses. The findings advance AI service failure research by developing a data-driven typology of failures and demonstrating that AI-mediated service breakdowns are not merely technical problems but relational threats that undermine consumer trust, satisfaction, and engagement with brands. However, because this research focuses on Chinese consumers and complaint data from a single national platform, future studies should examine whether the proposed failure typology and relational effects can be generalized across different countries, cultures, platforms, and regulatory environments.