The rapid expansion of social media has created multilingual and transnational communication spaces where users actively mix languages to express identity, build social connections, and participate in global digital culture. In Ambon, Indonesia , a multilingual region where Indonesian, Ambonese Malay, and English coexist , code-mixing is increasingly evident across platforms such as Facebook, Instagram, and TikTok. Yet, research on its linguistic patterns and social motivations in Eastern Indonesia remains scarce, with most studies centered on classrooms or major urban areas. This study analyzes the linguistic patterns of code-mixing among bilingual Ambonese social media users, identifies underlying motivations, and examines how local, national, and global languages interact to shape linguistic identity online. Adopting a descriptive–qualitative design, data were collected through purposive sampling from social media accounts of active bilingual users aged 17 and above. The corpus comprises status updates, comments, captions, and short video texts, supported by brief interviews. Analytical frameworks include Muysken’s typology and Myers-Scotton’s Matrix Language Frame Model. Findings show dominant use of word, phrase, clause, idiomatic, and hybrid insertions. Code-mixing serves to project cosmopolitan identity, enhance digital creativity, and negotiate belonging. This study contributes novel insights by situating code-mixing in Ambon’s multilingual context and integrating structural and functional analyses of online discourse.