For a page written in a non-English language you have three options for the slug: transliterate it to plain ASCII (schoene-gruesse-aus-koeln), keep the original script (schöne-grüße-aus-köln), or write the percent-encoded form of that script (sch%C3%B6ne-gr%C3%BC%C3%9Fe-aus-k%C3%B6ln). The last two are not really different: the percent-encoded form is how the Unicode slug travels over the wire, so the actual choice is ASCII or Unicode. Google's guidance is to use words in your audience's language in the URL, transliterated where that fits, and to percent-encode non-ASCII characters in your links. Either choice can work. What causes trouble is mixing rules, or letting Unicode create two spellings of one address.
What the Three Forms Look Like
| Slug written as | Example | Characters |
|---|---|---|
| ASCII transliteration | gemuese | 7 |
| Unicode (as shown to people) | gemüse | 6 |
| Percent-encoded UTF-8 (what is sent) | gem%C3%BCse | 11 |
| Japanese, Unicode | ペパーミント | 6 |
| Japanese, percent-encoded | %E3%83%9A%E3%83%91%E3%83%BC%E3%83%9F%E3%83%B3%E3%83%88 | 54 |
Russian привет-мир, percent-encoded | %D0%BF%D1%80%D0%B8%D0%B2%D0%B5%D1%82-%D0%BC%D0%B8%D1%80 | 55 |
URLs themselves can only contain a limited set of ASCII characters. RFC 3986 lists the "unreserved" ones (A–Z, a–z, 0–9, -, ., _, ~); anything else, including every non-ASCII letter, is sent as its UTF-8 bytes in %XX form. The newer IRI standard (RFC 3987) lets people see and type the Unicode form, and browsers convert it. Google's URL structure page shows https://example.com/gem%C3%BCse as recommended in an href and the raw gemüse as not recommended, even though the address bar usually displays the readable version.
So a Unicode slug is readable in the address bar and in search results, but whenever it is copied as raw text it grows: about twice as long for German, and five to nine times as long for Cyrillic or Japanese. That happens, for example, in some chat apps, logs, spreadsheets and analytics reports.
Option 1: Transliterate to ASCII
Transliteration maps each letter to a Latin spelling: ç to c, ş to s, é to e. For Latin-script languages this usually keeps the slug readable and short, and it never needs encoding.
The catch is that transliteration depends on the language, not just on the letter. German writes ö as oe when the umlaut is unavailable; Turkish readers expect ö to become plain o. Four widely used slug tools turn the same German title into four different slugs:
| Tool | Schöne Grüße aus Köln becomes |
|---|---|
WordPress, German locale (de_DE) | schoene-gruesse-aus-koeln |
| FindUtils URL Slug Generator | schone-grusse-aus-koln |
| WordPress, default locale | schone-gruse-aus-koln |
Django slugify() default | schone-grue-aus-koln |
WordPress's remove_accents() maps ß to s by default and only switches to oe, ue and ss for German locales (it has similar special cases for Danish ø to oe and å to aa). Django's slugify() decomposes the text and throws away anything that is not ASCII, so ß, which has no decomposition, simply disappears. The same method drops Ł, Ø and Æ: Django and the FindUtils generator both turn Łódź, Øresund and Æsir into odz-resund-and-sir.
Transliteration is also a poor fit for non-Latin scripts. A romanised Japanese or Arabic slug is rarely what local searchers type, and the tool you use may just delete the script: both Django's default and the FindUtils generator return an empty slug for Привет мир.
Pick transliteration when your audience writes in a Latin script, you want short shareable links, and you can fix the spelling rules for that language (for example oe/ue for German).
Option 2: Keep the Original Script
A Unicode slug matches the page title exactly and follows Google's "use your audience's language" advice. It is the usual choice for Cyrillic, Greek, Arabic, Hebrew, Chinese, Japanese, Korean and Thai sites. Django supports it with slugify(value, allow_unicode=True) (its own docs show 你好 World becoming 你好-world), and WordPress keeps non-Latin titles as percent-encoded UTF-8 slugs.
It comes with three traps that can quietly create two URLs for one page:
- Composed vs decomposed letters.
écan be stored as one code point (U+00E9) or aseplus a combining accent (U+0301). They look identical but encode differently:caf%C3%A9andcafe%CC%81. If one tool saves NFC and another NFD, you get two addresses. Normalise every slug to NFC (String.prototype.normalize('NFC')in JavaScript) before saving and comparing. - The Turkish dotted İ. In JavaScript,
"İstanbul".toLowerCase()returnsifollowed by a combining dot (U+0307), which encodes asi%CC%87stanbul, notistanbul. UsetoLocaleLowerCase('tr')for Turkish text (it also mapsIto dotlessı), or transliterate before lowercasing. - Case. URLs are case-sensitive, and many scripts have case too (
Ж/ж,Σ/σ). Lowercase consistently, the same way you would for English.
Pick Unicode when your audience reads a non-Latin script, or when transliteration would change the words they search for.
Option 3: Percent-Encode: the Form You Link With, Not a Separate Style
You should not design slugs as percent-encoded strings. Encoding is what your CMS, router or HTML should do automatically when it writes a Unicode slug into a link. Write links with the encoded form, as Google recommends, and make sure the encoding is consistent: uppercase hex (%C3%BC) and lowercase hex (%c3%bc) are equivalent under RFC 3986, but a mix in your sitemaps and canonical tags is a sign that two systems disagree. You can check what any slug encodes to with the URL Encoder Decoder, and split a full address into parts with the URL Parser.
A Quick Decision Table
| Your site | Suggested slug |
|---|---|
| English, or mixed audience with English slugs | ASCII |
| Latin-script language (Turkish, Spanish, Polish, German) | ASCII with language-correct transliteration, or Unicode; pick one and keep it |
| Non-Latin script (Russian, Greek, Arabic, Japanese, Chinese) | Unicode, normalised to NFC |
| Same page in several languages | One slug per language in its own folder (/de/, /ja/), linked with hreflang tags |
Whichever you choose, do not change slugs on published pages just to switch style. If you must, use permanent 301 or 308 redirects, as Google recommends for moved URLs.
Making ASCII Slugs with FindUtils
For Latin-script titles, the URL Slug Generator lowercases, transliterates and hyphenates as you type, and its bulk mode exports a whole list as CSV. It handles Turkish, French, Spanish and most Central European letters, but writes German ö/ü as o/u and removes letters like Ł and Ø, so review German, Scandinavian and Polish slugs by hand. The URL Slug Generator guide walks through every option with tested examples.
Need help with a project? The team behind FindUtils takes on freelance work through Upwork, including PHP bug fixes, a PHP 8 upgrade or new features for your website, for example cleaning up permalinks on a multilingual site.
Related Reading
Sources
- Google Search Central: URL structure best practices
- Google Search Central: Redirects and Google Search
- RFC 3986: URI Generic Syntax (sections 2.1 and 2.3)
- RFC 3987: Internationalized Resource Identifiers (IRIs)
- WordPress Developer Reference: remove_accents()
- Django documentation: django.utils.text.slugify
- MDN: String.prototype.normalize()
- MDN: String.prototype.toLocaleLowerCase()