Algospeak as algorithmic negotiation: multimodal strategies, platform governance, and the politics of visibility on TikTok

Abstract The expansion of automated content moderation has fundamentally reshaped communicative practices on social media platforms, particularly within short-form video environments, namely TikTok. In response to increasingly opaque systems of algorithmic governance, users have developed algospeak : a repertoire of coded linguistic and multimodal strategies that enable the expression of sensitive content while mitigating the risk of algorithmic suppression. This study examines algospeak as a form of algorithmic negotiation, through which users actively recalibrate meaning-making practices under conditions of computational visibility control. Adopting a mixed-methods design that combines multimodal discourse analysis (MDA), digital ethnography, and corpus-assisted semantic exploration, the study analyses a dataset of 550 TikTok posts. The findings demonstrate that algospeak operates as a multimodal semiotic ecology, in which meaning is systematically redistributed across textual, visual, auditory, and performative channels. Strategies such as orthographic manipulation, emoji encoding, euphemistic reformulation, translanguaging, and multimodal masking function not as isolated linguistic innovations but as coordinated practices that sustain a dual intelligibility. Algospeak emerges as a creative self-censorship and an adaptive response to platform governance, reflecting growing forms of algorithmic awareness and anticipatory self-regulation among users.