Pseudo-localization
A fake locale that turns English into accented, padded look-alike text so you catch hardcoded strings and layout breaks before real translation starts.
See it
What it is
A pseudo-locale machine-transforms your source strings instead of translating them, and it does three jobs at once. Accenting every character ('Save' becomes 'Şåvé') proves your resource pipeline, fonts, and rendering survive non-ASCII. Padding each string by 30% to 50% proves the layout tolerates text expansion. Wrapping it in markers like [[ ]] proves the string actually came from the message catalog and shows you exactly where it gets truncated.
Anything still rendering as plain unaccented English is hardcoded, which makes this the cheapest hardcoded-string detector there is. Android and Chrome ship this idea as real locale codes: en-XA for accented and padded, ar-XB for a mirrored right-to-left pseudo-locale that flags layout that will not flip. Reach for it the day you finish string extraction, long before the first vendor quote.
Gotcha: a clean pseudo-locale pass only proves mechanical readiness. It says nothing about plural categories, grammatical gender, or whether the sentence makes sense once a real translator reorders it. It also proves nothing about input handling, because transforming catalog strings never touches the path where a user types Unicode text. Test that separately with fixtures that enter, store, retrieve, and edit non-ASCII input. Also gate it behind a dev flag or env var; a pseudo build reaching production looks like a hack, not a test.
Ask AI for it
Add a pseudo-localization locale to this project's i18n setup. Generate it at build time from the source catalog: map ASCII letters to accented look-alikes (a to å, e to é, o to ö, and so on), pad every string to roughly 140% of its original length with repeated vowels, and wrap each value in [[ and ]] brackets. Register it as locale 'en-XA', expose it behind an env flag so it never ships to production, and add a second pass that renders the pseudo-locale in right-to-left mode.