Loneliness Predicts Intimacy in 277 AI Companion Users, Shaped by Attachment and Age

Researchers are increasingly investigating the potential of artificial intelligence (AI) companions to alleviate loneliness, yet understanding of how individual differences influence these relationships remains limited. Raffaele Ciriello, Uri Gal, and Ofir Turel, from the University of Sydney and University of Melbourne, present new findings examining the complex interplay between loneliness, attachment styles, age, and intimacy with AI companions. Their collaborative work, drawing on a hermeneutic literature review and a survey of 277 active users, demonstrates that AI companions are not a universal solution to loneliness. The study reveals a nuanced pattern where the relationship between loneliness and intimacy is significantly moderated by attachment insecurity and conditioned by age, highlighting the importance of considering psychological dispositions and demographic factors when designing and deploying these technologies. These findings are significant as they clarify who benefits most from AI companionship and raise crucial ethical considerations regarding the potential for commercial exploitation of vulnerable users.

This work challenges the simplistic notion that AI companions offer a universal solution, revealing instead that the development of intimacy with these agents is a complex ‘sociotechnical process’ shaped by pre-existing psychological traits and demographic factors.

A study involving 277 active users of AI companions demonstrates a surprising pattern: loneliness does not consistently translate into increased intimacy. The research highlights a paradoxical effect where securely attached individuals, those generally comfortable with close relationships, actually exhibit reduced intimacy when experiencing loneliness, while those with avoidant or ambivalent attachment styles report increased intimacy with AI companions.

Attachment theory, a psychological framework explaining how early childhood relationships shape adult bonding patterns, provides a lens for understanding these divergent responses. Drawing on a hermeneutic literature review and a detailed survey, the researchers developed a model demonstrating that loneliness predicts intimacy, but this relationship is significantly moderated by attachment insecurity and conditioned by age.

This work clarifies who is most likely to form intimate bonds with AI companions, and importantly, raises ethical concerns about commercial models that might exploit user vulnerabilities. By moving beyond broad claims about the benefits of AI companionship, this study offers a critical assessment of artificial intimacy and its potential consequences.

Loneliness exhibited a complex relationship with intimacy formation amongst users of AI companions, with patterns significantly shaped by both attachment style and age. Analysis of data collected from 277 active users revealed a paradoxical effect for securely attached individuals, where increased loneliness corresponded with reduced reported intimacy levels.

Conversely, both avoidant and ambivalent attachment styles demonstrated increased intimacy alongside heightened loneliness, suggesting these users may turn to AI companions to fulfil relational needs unmet in their existing connections. Anxious attachment showed mixed effects, indicating a more nuanced interplay between loneliness and intimacy. Further differentiating these patterns, older adults consistently reported higher levels of intimacy even when experiencing relatively low levels of loneliness, suggesting a potential benefit for this demographic, possibly due to reduced social networks and a greater appreciation for companionship regardless of attachment security.

The study’s model demonstrates that loneliness alone is not a straightforward predictor of intimacy with AI companions; specifically, the research highlights that the strength and direction of this relationship are contingent upon pre-existing psychological dispositions and demographic factors. These findings challenge the notion of AI companions as a universal solution to loneliness, instead positioning artificial intimacy as a sociotechnical process.

The data underscores the importance of considering individual vulnerabilities when designing and deploying these technologies, as they may disproportionately impact certain user groups. A hermeneutic literature review initiated this work, moving beyond systematic approaches to foster iterative interpretation and conceptual refinement within the emerging field of AI companionship.

This review, conducted across Scopus, Google Scholar, and key disciplinary outlets, interrogated assumptions present in both academic research and commercial discourse surrounding these technologies. Search terms encompassing loneliness, intimacy, attachment, and related keywords were employed, deliberately including peer-reviewed conference papers and high-quality media analyses to capture the field’s rapid evolution.

The goal was not exhaustive coverage, but rather a nuanced understanding of the interplay between psychological dispositions and the adoption of AI companions. Following the literature review, a quantitative survey was distributed to 277 active users of AI companion applications to examine the relationships between loneliness, intimacy, and attachment styles.

Participants were recruited through online platforms and social media channels frequented by AI companion users, ensuring a diverse sample actively engaging with these technologies. The survey instrument incorporated established scales for measuring loneliness, assessing both emotional and social dimensions, and attachment insecurity, categorising individuals along avoidant, ambivalent, and secure attachment dimensions.

Intimacy was operationalised as a self-reported measure of perceived relational engagement with the AI companion, focusing on affective disclosure and perceived responsiveness, mirroring key components of human intimacy. Age and demographic data were also collected to account for potential cohort effects and individual differences. This mixed-methods approach, combining interpretive literature review with quantitative survey data, allowed for a comprehensive exploration of the complex sociotechnical processes shaping intimacy with AI companions, moving beyond simplistic assumptions of universal benefit or detriment.

The relentless pursuit of connection in the digital age has yielded a curious paradox: increasingly sophisticated AI companions marketed as antidotes to loneliness, yet deployed with little understanding of who actually benefits. This research moves beyond simplistic narratives of technological salvation, revealing that the relationship between loneliness and intimacy with AI isn’t straightforward, but a nuanced interaction shaped by pre-existing psychological traits and life stage.

For years, studies have highlighted the growing epidemic of loneliness, particularly amongst younger and older demographics, but the assumption that a digital companion could simply ‘fill the void’ overlooked the complex reasons people experience social disconnection. What’s particularly compelling is the finding that attachment styles significantly moderate the effect; securely attached individuals seem less likely to seek intimacy with AI when lonely, suggesting a robust existing support network or a healthier coping mechanism.

Conversely, those with avoidant or ambivalent attachment styles appear more readily drawn to these digital relationships, raising ethical concerns about exploitation of vulnerability. The higher reported intimacy among older adults, even at lower loneliness levels, hints at a different dynamic, perhaps a pragmatic acceptance of companionship in the face of limited social opportunities.

However, the cross-sectional nature of the data prevents establishing causality. Do these individuals turn to AI because they are lonely, or does the AI interaction itself alter their experience of loneliness? Future research must employ longitudinal studies to unravel this temporal relationship. Moreover, the study prompts a broader question: what constitutes ‘intimacy’ with an artificial entity, and how does that differ from human connection.