If your audience spans languages, tags become a real decision rather than an afterthought. Getting it wrong fragments your reach across two half-populations.
Tag in the language of the audience, not the creator
The obvious point that gets missed: tags should match how your viewers search, not what language you think in. A Berlin-based account whose audience is mostly American should tag in English. A Miami account serving a Spanish-speaking community should tag in Spanish, regardless of what language the caption uses.
Mixing languages within one set
Generally a mistake. A set that is half English and half Spanish reaches a diluted slice of both audiences, and the classification signal gets muddier. If you genuinely serve both, the better structure is separate posts, or a caption in both languages with tags weighted toward the primary market for that specific post.
Where mixing does work
Bilingual communities that genuinely code-switch — Spanglish content, Hinglish content, Arabic-English professional content — have their own established mixed tags. If your audience is that community rather than two separate audiences, use their vocabulary.
Accents and special characters
Platforms handle these inconsistently. Some strip accents, some treat accented and unaccented forms as separate tags, and users type both. Where it matters, include both forms — the pool is usually small enough to afford the slot.
Do not translate a tag literally
The direct translation of a popular English tag is often not the tag native speakers actually use. Communities develop their own vocabulary. The only reliable method is to search the topic in the target language and read what tags the local accounts use.
Using the language setting here
Selecting a language in the generator changes the connective vocabulary used to build compound tags, so results read like a native speaker wrote them rather than like English with a foreign word attached. For a fully localised set, type the keyword in the target language too.