ChatGPT's ability to realistically mimic human conversation and its high level of ability to handle linguistic ambiguity opens new and exciting avenues in language learning. Building upon the technical affordances of ChatGPT, this study explores the perceptions of educational affordances when incorporating ChatGPT across languages discussed by language communities on YouTube and identifies best practices for its effective use in language education. Through inductive content analysis, this study discussed 18 languages categorized into four groups: (1) romanized languages with high resources, (2) non-romanized languages with high resources, (3) languages with medium resources, and (4) less frequently used languages. The findings reveal consensus that (a) ChatGPT is a valuable and remarkable tool for language teaching and, (b) learning and it cannot fully replace teachers, as humor, wit, and sympathy cannot be programmed. Two potentially significant issues or two gaps were identified and discussed: namely, the learning optimization gap and the knowledge comprehension gap.