Can You Use a Non-English Passphrase? Yes, With One Test First
Can you use a non-English passphrase? Short answer: yes, on most modern sites, and I think it is a perfectly reasonable choice if that language is the one you actually type in every day.
My grandmother typed in two languages and never once thought of it as a security feature. She just used the keyboard she knew. That is the right way to think about this question. A passphrase built from words you type without thinking is stronger in practice than a longer English phrase you have to look up, misspell, and eventually reset.
The standards are on your side here. OWASP ASVS requirement 2.1.4 says Unicode characters should be permitted in passwords, and RFC 8265 defines a password profile built around the full Unicode FreeformClass. In plain terms, a well built site should accept your words whether they are in Arabic script, Cyrillic, Devanagari, or Japanese kanji. A site that rejects them is usually running an old input filter, not enforcing a real security rule.
That gap between what should happen and what does happen is where the trouble lives. I would not set a non-English passphrase on an account I cannot afford to lose without running one test first. Set the passphrase, log out completely, and log back in from a different device. Then try it from your phone, where the keyboard layout and autocorrect behave differently. If it works in all three places, you have a passphrase. If it fails on the phone, you found the problem before it locked you out at the worst possible moment.
Can you use a non-English passphrase on every site?
No, and the failures are rarely announced. Some older systems store passwords in a single byte encoding and silently drop or convert characters they do not recognize. The result is a passphrase that looks right on screen and fails every time you type it, because the stored version is not the version you typed. Banking portals, corporate VPNs, and pre-boot disk encryption prompts are the usual suspects. Modern web apps built on UTF-8 mostly handle this fine, but a login screen that only exists before the operating system fully loads may not.
Normalization is the quieter trap. Unicode lets the same visible character be stored in more than one byte sequence. A letter with an accent can be one precomposed character or a base letter plus a combining mark. They look identical. A site that normalizes on registration but not on login, or the other way around, will reject a correct passphrase. This is why the test in the previous paragraph matters more than any promise in a help article. Your eyes cannot see normalization differences, so only a real login attempt can catch them.
There is also the keyboard problem nobody talks about until they travel. A passphrase in a script you type fluently at home becomes a puzzle on a borrowed laptop in a hotel lobby, or on a phone whose keyboard layout switched after an update. I keep a simple rule for this. If I cannot type the passphrase on a plain US keyboard layout using only characters I can produce without installing anything, it does not go on an account I might need while traveling. That rule costs me some creativity and has saved me at least one lockout I can remember clearly.
What I would actually do
Use words from the language you think in, drawn at random, not a sentence from a book or a famous lyric. Four or five random words in any script carry the same math as four or five random English words, because the security comes from randomness, not from the alphabet. A phrase everyone in your language knows is a dictionary entry, and cracking tools carry word lists in many languages now. Randomness is the whole trick. The script is just the wrapping.
Keep the passphrase in a password manager that handles Unicode cleanly, and let the manager type it for you on the devices where typing is awkward. Test the login on each device once, write down nothing on paper, and keep a recovery method that does not depend on the same passphrase. One question I still leave open, honestly, is whether I would put a non-English passphrase on full disk encryption. The standards allow it and modern UEFI prompts handle it, but the recovery story on a machine that will not boot is thinner, and I have not settled that tradeoff for my own laptop yet.
Frequently asked questions
Can you use a non-English passphrase on a website?
Yes, most modern websites built on UTF-8 accept Unicode passphrases. OWASP guidance says Unicode characters should be permitted, but older systems may still reject or mangle them, so test the login on every device you use.
Does a non-English passphrase make my password stronger?
Only if the words are random. Four random words in any language carry similar strength to four random English words. A famous quote or common phrase in any language is weak, because cracking tools carry word lists in many languages.
Why does my Unicode passphrase work on one device but not another?
The usual causes are keyboard layout differences and Unicode normalization, where the same visible character is stored as different bytes. Log out and test on each device before you rely on the passphrase.
Should I use a non-English passphrase for disk encryption?
It can work on modern UEFI systems, but the recovery story is thinner if the machine will not boot and the pre-boot keyboard behaves differently. For disk encryption, test the passphrase at the boot prompt before you commit to it.
Are emoji passphrases a good idea?
Emoji are Unicode characters, so they can work where a site accepts them, but they are hard to type on many keyboards and easy to mistype. Random words in a language you type fluently are usually the safer choice.
Related reading: How Many Words Does Your Passphrase Need? Entropy Math, Worked Out · Passphrase vs Password: What Is the Difference? · Is It Safe to Use an Online Password Generator? · Words to Avoid in a Passphrase