v151 · HTML · i18n
Locale & Script Matrix
Eight scripts and writing systems, each with an autocorrect="on" and an autocorrect="off" input. Type native text (or use the sample prompts) and see how the OS handles autocorrection differently per locale. Toggle all inputs globally to compare baseline vs off behaviour.
IME note — For CJK and Arabic scripts, autocorrect interacts with IME composition. Type as you normally would; the difference is most visible on mobile devices where the OS autocorrect engine is more aggressive. On desktop, effects are subtler but still observable with spell-correctable words.
Global override:
| Language / script | autocorrect=on | autocorrect=off | Sample to type | IME / platform mediation | Autocorrect notes |
|---|
Per-locale behaviour notes
Latin scripts (en, fr, de, es, pt)
Autocorrect is most active here. Common words get silently replaced — "teh" → "the", "recieve" → "receive". For prose inputs this is helpful; for code, search, or proper nouns it's harmful. The
autocorrect attribute gives precise per-field control.Japanese (CJK) — ja
IME handles kana → kanji conversion. OS autocorrect rarely fires during composition but can substitute words after commit. Technical fields (product codes, IDs) should use
autocorrect="off" to prevent unexpected substitutions. IME and autocorrect are separate layers.Chinese — zh-Hant / zh-Hans
Similar to Japanese: IME handles romanisation-to-character conversion. Autocorrect mainly affects the romanised input phase. For Traditional/Simplified distinction, autocorrect could incorrectly replace characters; disable for formal document editing.
Arabic / Hebrew (RTL)
Both are RTL scripts; inputs must use
dir="auto" or explicit dir="rtl". Arabic OS keyboards include autocorrect for common diacritics and root-form variations. Hebrew autocorrect fills in nikud. Both should be off for technical or proper-noun fields.Devanagari — hi
Typed via phonetic romanisation → Devanagari transliteration engines. Autocorrect can interfere with the phonetic input by "correcting" the romanised syllables before they're converted. Disable for form fields where exact transliteration matters.
Korean (Hangul) — ko
Hangul is typed syllable-by-syllable via IME. Autocorrect at the OS level is relatively uncommon for Korean prose but active for mixed Korean/English (Konglish). Username and code fields should still use
autocorrect="off".references
implementation reference
Need the exact API surface, compatibility boundaries, errors, lifecycle, and source links? Read the matching gendn reference ↗