Emotional dependence on AI chatbots: Development and Validation of the AIED Scale

As large language model (LLM) chatbots become increasingly embedded in adolescents’ daily lives, concerns have emerged about users’ emotional reliance on these systems. Existing measures of AI chatbot dependence primarily assess instrumental or problem-focused use and do not adequately capture emotional dependence. This study developed and validated the AI Chatbot Emotional Dependence (AIED) Scale, a brief measure of adolescents’ tendency to rely on AI chatbots for comfort, emotional relief, and perceived understanding. A total of 5,855 Chinese adolescents with complete data were randomly divided into an exploratory subsample (n=2,927) and a confirmatory subsample (n=2,928). Exploratory factor analysis supported a one-factor, five-item solution after removal of one poorly performing item. The retained factor explained 69.07% of the variance, with loadings ranging from .778 to .864. Confirmatory factor analysis supported the unidimensional structure, with good overall fit (CFI=.98, TLI=.95, SRMR=.02). The scale showed good internal consistency (α=.89), strong convergent validity, and a substantial positive correlation with a general AI chatbot dependence measure (r=.67,p<.001). Multi-group analyses supported configural, metric, and scalar invariance across sex and educational stage. The AIED is a brief, reliable, and valid instrument for assessing adolescents’ emotional dependence on AI chatbots.