The most vulnerable are prone to use AI therapists: The role of attachment, epistemic trust, and mental health symptoms in acceptance of digital mental health interventions

OBJECTIVE: Digital mental health interventions (DMHIs) offer scalable solutions to traditional barriers in mental healthcare. Acceptability of DMHIs might differ among patients with differing level of symptoms and relational difficulties. This study investigated how a person's psychological profile, including their level of mental health symptoms, epistemic trust, and attachment security, predicts their acceptance of AI-based interventions (avatars, chatbots) versus human-delivered teletherapy. METHOD: Using cluster analysis on survey data from 1,612 (potential) patients and clinicians recruited via Prolifc, we identified three distinct psychological profiles: "Avoidant-not trusting," "Secure-trusting-healthy," and "Young-anxious-ambivalent-symptomatic." RESULTS: Results showed a paradoxical pattern: Individuals in the two most vulnerable clusters ("Avoidant-not trusting" and "Young-anxious-ambivalent-symptomatic") demonstrated significantly higher acceptance of AI-based DMHIs. In contrast, the psychologically healthier "Secure-trusting-healthy" cluster showed the lowest AI-based DMHI acceptance and the highest acceptance of teletherapy. Also, having therapy experience as provider or patient was associated with lower AI acceptance in general. CONCLUSION: These findings suggest that AI-based interventions may be uniquely suited to reach individuals who struggle with human-to-human therapeutic relationships, thereby serving as a critical entry point to care for those who need it most. However, those vulnerable people might also miss out on the relational learning that takes place in human-based therapies.