Analisis Modifikasi Ejaan Kontekstual sebagai Strategi Netizen Menghindari Sensor Otomatis di Media Sosial TikTok

Abstract

The strictness of automated content moderation systems (keyword filtering) on the TikTok platform often triggers the shadowban phenomenon against public discourse containing legal information or sensationalized crime news. This reality drives Indonesian netizens to carry out linguistic resistance through contextual spelling modifications. This study aims to analyze the forms of spelling modifications used by netizens on the TikTok platform and to test their effectiveness in circumventing the platform's automated censorship system. Using a qualitative descriptive approach within the framework of cyber sociolinguistics, textual data were collected through passive participatory observation and the note-taking technique from comment sections and video captions on TikTok Indonesia throughout the period of January to March 2026. Data analysis was conducted interactively following the Miles, Huberman, and Saldaña model. The results demonstrate three main typologies of spelling modification: (1) alphanumeric graphemic substitution based on visual resemblance (such as encoding violent words into T3W4S and D1T3MB4K), (2) morphological segmentation and tactical acronymization (bundir, pekob), and (3) semantic transformation based on visual euphemisms (saus tomat, bercocok tanam). These findings prove the technical limitations of TikTok's Natural Language Processing (NLP) technology in detecting rigid visual orthographic distortions. On the contrary, human readers remain cognitively capable of reconstructing messages instantly based on the pragmatic context of the speech. This study concludes that contextual spelling modification functions effectively as a content survival mechanism that confirms the arbitrary, dynamic, and adaptive nature of human language in the face of artificial intelligence dominance.