Gaussian Graphical Mixture Model (GGMM)-based clustering represents an appealing framework to jointly uncover natural clusters in the data while accounting for complex dependence structure among variables. To explore how digital self-image relates to different patterns of social media use, psychological traits, and social contexts of development, we leverage this approach to profile the online behaviours of adolescents actively using social media. Our goal is to identify those most vulnerable to digital technologies, while simultaneously targeting, within a network framework, pivotal constructs on which to focus in tailoring intervention programs promoting digital well-being. A four-cluster solution emerged as the best, considering different estimation algorithms and penalty functions. The identified clusters displayed distinct patterns in online self-image management, psychological well-being, and perceived social support, while differing in network structure, not only in terms of density but also centrality measures. Interestingly, self-esteem related to real interpersonal relationships emerged as pivotal node in most of the cluster-specific networks.