Why it matters
- A crawler reads every difference in an address as another page.
قهوةandقَهوةare two addresses, and so are/Coffeeand/coffee: the same page can end up indexed twice, or the second spelling can answer with a 404. - People type an address, copy it from a message and share it in chats. A slug with marks, Eastern digits or capital letters is the one they get wrong; an underscore is not read as a word break the way a hyphen is, and Google recommends hyphens.
- A very long slug is encoded for sharing (each Arabic letter becomes six characters), so it grows into a line nobody can read in a message or a result.
- Google reads Arabic in addresses, so a clean Arabic slug is fine. The slug generator writes one that is Arabic, short and stable.
Example
<link rel="canonical" href="https://shop.example/products/%D9%82%D9%8E%D9%87%D9%88%D8%A9_%D8%B9%D8%B1%D8%A8%D9%8A%D8%A9-%D9%A2%D9%A0%D9%A2%D9%A6" />
<link rel="canonical" href="https://shop.example/products/%D9%82%D9%87%D9%88%D8%A9-%D8%B9%D8%B1%D8%A8%D9%8A%D8%A9-2026" />
How to fix
Write the slug from the title with words in one spelling, joined by single hyphens:
html
<link rel="canonical" href="https://shop.example/products/قهوة-عربية-2026" />- Take the marks and the tatweel out of the slug (
مَرحباbecomesمرحبا) and write digits as0to9. - Use a hyphen between words, lowercase Latin letters, and no spaces, underscores or punctuation.
- Keep it to a few words, and let the product's name carry the rest.
- Do not change a slug that is already ranking only to tidy it: set up a 301 redirect from the old address to the new one first.
- Platforms write the slug from the title: on Salla, Zid and Shopify the slug is a field of the product or the page, next to its SEO title.
How we detect
- We take the page's address and the canonical it names (the canonical only when its path differs), and read each path segment as people read it, with the percent-encoding undone.
- A segment is flagged for marks if it has harakat, the superscript alef or tatweel; for digits if it has Arabic-Indic or Persian digits; for separators if it has an underscore, a space, a plus sign, a percent sign left after decoding, a capital Latin letter, a double hyphen or punctuation such as a comma, a quote or a bracket.
- The last segment is flagged as long past 100 characters, counted as people read it, not as it is encoded. This threshold is Arablyzer's: Google gives no number.
- The home page, and a page whose address and canonical are both the root, are not judged.
We judge the address, not the title: a title with marks is fine as a title. Query strings are left to the URL-parameters check.