In recently published studies on cyberbullying, students are frequently categorized into distinct (cyber)bully and (cyber)victim clusters based on theoretical assumptions and arbitrary cut-off scores adapted from traditional bullying research. The present study identified involvement classes empirically using latent class analysis (LCA), to compare the classification of cyber- and traditional bullying and to compare LCA and the conventional approach. Participants were 6,260 students (M = 14.8 years, SD = 1.6; 49.1% male) from six European countries. LCA resulted in three classes for cyberbullying and four classes for traditional bullying. Cyber- and traditional bullying differed from each other, as did LCA and the conventional approach. Country, age, and gender differences were found. Implications for the field of traditional and cyberbullying research are discussed.