ABSTRACT AI companion applications are transforming how people work, interact, and even seek companionship by increasingly taking on roles once reserved for humans. While AI companions are being rapidly adopted for comfort and connection, research on the dark side of such interactions remains underexplored in consumer behaviour research. Drawing on the stressor‐strain‐outcome (SSO) framework, the current study investigates user‐reported experiences with Replika, a widely used AI companion application, to uncover dominant themes and behavioural patterns, revealing how AI companions can negatively impact consumers. To this end, we utilised the deep contextual understanding of BERT (Bidirectional Encoder Representations from Transformers)‐based models and analysed 157,077 reviews from the Google Play Store. To enhance our insights, an empirical study is conducted that links the consumer experiences with quantifiable behavioural outcomes, using the ratings as a proxy for consumer satisfaction. To the best of our knowledge, this is among the earliest studies to combine large‐scale NLP and consumer satisfaction modelling in the context of AI companion apps. Our findings reveal critical consumer concerns, including technical issues, monetisation concerns, ethical risks, and human–AI relationship challenges. Additionally, we found significant behavioural and psychological impacts, including technological dependence, social disconnection, and emotional dependence. Furthermore, by extending the SSO framework in the context of AI companionship, we establish links between negative experiential themes and behavioural outcomes. Our research contributes to the growing body of literature on the ‘dark side’ of AI companionship, offering theoretical insights into consumer behaviour literature and practical guidance for designing ethical and consumer‐centric AI companion applications.
Gautam et al. (Mon,) studied this question.