•  
  •  
 
Journal of Counseling and Psychology

Abstract

The rapid growth of large language models (LLMs) raises questions for counselors regarding when such tools are used adaptively versus when they foster dependency. Guided by attachment theory and a dual-dimension view of AI reliance, this study examined whether adult attachment styles predict instrumental (task/decision support) and relational (companion-like) dependency on LLMs, whether spiritual engagement relates to these dependencies, and whether levels differ by age. Faculty, staff, and students who had used an LLM completed the LLM-D12 and the ECR-R (N = 515). Analyses showed that instrumental dependency on LLMs was highest among individuals high in avoidance, with modest additional elevation when anxiety is also high. Relational dependency on LLMs increased progressively from the secure to the anxious-avoidant style, with the highest scores among individuals elevated on both dimensions. Multivariate analyses indicated that frequent “talking with God” (multiple times/day) was linked to lower instrumental reliance for confidence/support/decisions and to lower endorsement of LLMs as relational tools or companions. Age was not associated with differences in instrumental dependency but was reliably associated with differences in relational dependency: with younger adults reporting higher relational dependency. Findings suggest that avoidance-based attachment style primarily predicts task-focused reliance on LLMs, while combined attachment insecurity relates to companion-like engagement. For counselors and counselor educators, results underscore the value of assessing client motives for AI use and targeting relational overreliance—especially in younger adults—through psychoeducation and skills that strengthen human connection.

Download

Share

COinS