Understanding User-Generated Content and Communities in Decentralized Web3 Networks: An Empirical Study on Nostr

Decentralized social protocols such as Nostr introduce a new paradigm for user-generated content (UGC) in the Web3 era, where content production, dissemination, and reward mechanisms operate without centralized governance. This paper presents one of the first large-scale empirical analyses of Nostr, based on 22.3 million user events collected from four major publicly accessible relays. Guided by three research questions, we examine (1) the temporal and spatial distribution of user participation, (2) the structural characteristics of decentralized UGC networks, and (3) thematic and incentive patterns in content creation and Zap-based rewards. Our analysis shows rapid growth followed by long-tail stabilization, while the interaction network remains highly modular and loosely connected, indicating fragmented yet persistent communities. Embedding-based clustering of textual posts identifies ten clusters on several topics: technical discussions, ideological debates, personal expression, community coordination, and media sharing, highlighting a hybrid ecosystem of social and technical discourse. We further find that knowledge-oriented content in Clusters 1 and 5 receives higher Zap engagement, suggesting the socialization of a primarily technical infrastructure. These findings advance the understanding of decentralized multimedia ecosystems by linking network decentralization with observed participation and engagement patterns in the absence of centralized moderation.