This article looks at the semiotics of humorous self-censorship. To this end, selected examples from a corpus of YouTube commentary videos and their respective comment sections are presented and discussed. On the one hand, the analysis focuses on structural aspects of humorous self-censorship signs from various modes (written, spoken, images, emojis, etc.) in their multimodal interplay. On the other hand, we analyze socio-semiotic aspects of our examples: their anchoring in specific speech communities, marked by background knowledge and shared communicative practices. The analysis shows that (1) structural manipulations of the spelling and/or phonetic shape of lexical items and of images etc. serve to secure both the understanding of the censored item as well as plausible deniability, while generating potentially humorous incongruities, (2) various positions in the participation process are exploited to participate in this process, including trigger warnings and other metacommunicative actions and levels, (3) humorous self-censorship serves a number of social-semiotic functions, such as the negotiation of group norms of sayability, the expression of group solidarity, and – importantly – entertainment.