Bilingual SEO: the Two Tags Your Arabic Pages Depend On

You can translate every page, publish both versions, and still have a search engine index only one of them. Nothing errors. The pages load. They are simply being told, in the markup, that they are duplicates of each other.

A canonical tag names the address a page wants credited. On a bilingual site the rule is short: a page that exists in two real languages is its own canonical in each. The Arabic page names the Arabic URL. The English page names the English one.

The common failure is a single canonical generated from a language-neutral path, so every Arabic page names its English twin. That tells a crawler the Arabic version is a duplicate and should be dropped — which it then is.

There is one case where pointing back is correct: a page that is only translated in its chrome. If the copy is still English and only the layout flipped, that address is a duplicate, and consolidating it is the honest signal.

Canonical says which URL counts. hreflang says which language each one serves, and it only works in reciprocal pairs — the English page points at the Arabic and the Arabic points back. A one-way annotation is ignored.

• Annotate only pages that genuinely exist in both languages
• Make every pair reciprocal, and include each page's reference to itself
• Add x-default for the version a reader lands on when nothing matches

A sitemap listing only the English URLs quietly says the other language is not for indexing. List both, and carry the same alternate set on each entry.

The three have to tell one story. Any two agreeing and a third contradicting them is resolved by the crawler, not by you.

Worth checking on a site you already run: open an Arabic page, view source, and read the canonical. If it names the English URL, that is the whole problem.