Artificial intelligence (AI)-powered chatbots are becoming embedded in everyday life, offering users not only practical assistance but also forms of emotional companionship. Yet the prevalence of emotional attachment to chatbots and its formation remain insufficiently understood. Drawing on a cross-sectional survey conducted in Germany, China, South Africa, and the United States, this study examines the extent, antecedents, and consequences of emotional attachment to chatbots. We propose a dynamic interaction framework of emotional attachment, which emphasizes the interplay between perceived emotional support, social connectedness, and the depth and frequency of chatbot use. The findings reveal that emotional attachment is strongly associated with perceived emotional support, including reduced loneliness, freedom from judgment, and a sense of privacy. Frequent and deep use of chatbots further reinforces emotional bonds, while surprisingly, larger social networks also correlate with stronger attachment. Additionally, regression analysis suggests that emotional attachment translates into user dependence. These results illuminate the growing emotional entanglement between humans and AI systems, raising new questions about digital well-being, human-machine intimacy, and the governance of affective technologies in an era of AI-driven social transformation.