This study aims at investigating how loanwords from Japanese and Korean are used in informal written Cantonese media discourse, including print and social media. Data from these media were collected from designated websites for 15?min every other day over a two-week period. The results show that loanwords from Korean, being written in a phonographic script "hangul" ([foreign characters omitted]), are rendered into written Cantonese typically through phonetic adaptation using Chinese morpho-syllables, while their Chinese-specific morphographic meanings are ignored. By contrast, lexical items from Japanese written in "kanji" tend to be borrowed directly through graphic borrowing, paying no regard to their Japanese pronunciation. Japanese being written with mixed scripts, "kanji" and two "kana" syllabaries, graphic borrowing from "hiragana" or "katakana" is rare, with the Japanese grammatical particle [foreign character omitted] being a notable exception. We conclude that lexical items written in a phonographic script tend to be rendered into written Cantonese phonetically, while those written in character-based "hànzì" are borrowed directly through graphic borrowing but assigned Cantonese pronunciation. In informal interaction between Cantonese-dominant Hongkongers, colloquial written Cantonese relies on the affordance of script mixing for its vitality, in print as much as in internet-mediated social media discourse.