The emergence of generative artificial intelligence (AI) has opened new dimensions in how humans form emotional bonds with technology. As AI becomes increasingly integrated into daily life, interactions with technology are evolving beyond functionality toward emotional engagement. In a recent survey, 62% of U.S. adults reported interacting with AI several times a week (Kennedy et al., 2025). Similarly, over 85% of U.S. high school students use AI, with 42% seeking emotional or mental health support and 19% reporting romantic involvement, particularly among those using AI for diverse purposes (Laird et al., 2025). In line with these trends, a growing number of individuals engage with AI platforms as friends, companions, therapists, or romantic partners (Robb & Mann, 2025; Tran et al., 2025). Although research increasingly examines human-AI emotional attachment through the lens of Attachment Theory (Heng & Zhang, 2025; Hu et al., 2025), findings remain fragmented, underscoring the need for theoretical integration and synthesis. Thus, we conducted a scoping review to examine theoretical approaches to human-AI relationships, explore how attachment has been conceptualized and measured, and identify key trends and research gaps. Following the PRISMA-ScR guidelines (Tricco et al., 2018), we conducted a comprehensive search of peer-reviewed articles published between 2020 and 2025 in PsycINFO, Web of Science, and Scopus. Inclusion criteria comprised empirical and theoretical studies examining human-AI relationships through attachment frameworks, particularly those involving generative AI, AI companions, or therapeutic AI systems. 959 articles were identified, and after screening and exclusions, 76 met the eligibility criteria. Preliminary findings demonstrated that most studies employed attachment theory as their theoretical framework. Furthermore, factors that promote or hinder the strengthening of human-AI relationships (e.g., anthropomorphism, perceived empathy, loneliness, and trust) were identified. These findings highlight the need for cohesive frameworks to guide future theoretical and empirical research on human-AI attachment.