Canonical address.
A canonical address is the markup declaring which page is the original when the same or very similar content is reachable from more than one address.
On multilingual sites this declaration comes in at a critical point and is often misused: every language version must point canonically to ITSELF. Translations are not copies of one another but separate pages.
The wrong use is this: the German page’s canonical points to the Turkish page. The result is usually that the German version never enters the index — the engine counts it a copy and discards it. Everything hreflang says goes to waste.
The two pieces of markup talk to each other, and when they conflict the canonical wins. So when auditing hreflang the canonicals have to be checked too — the two can each be right on their own and wrong together.
- Every language version points canonically to itself — without exception.
- Address variants are gathered into one form — trailing slash, capitalisation, parameters.
- The canonical has to be an absolute address — a relative path carries risk.
- A canonical can point to itself — and should.
On this site every generated page points canonically to itself, and that was verified by measurement; not a single page is missing or points elsewhere. Address variants — forms without the trailing slash and with a file name — are redirected to a single form.
The sources of the same content being reachable from several addresses are well known: forms with and without a trailing slash, capitalisation variants, tracking parameters, and combinations of sorting or filtering.
The declaration has to be an absolute address and can point to itself — indeed it should. A page pointing canonically to itself declares the original form whichever address variant the visitor arrived from.
On multilingual sites it has to be audited together with hreflang: the two can each be right on their own and wrong together, and when they conflict the canonical wins.
- Every language version points canonically to itself — without exception.
- It has to be an absolute address — a relative path carries risk.
- It can point to itself, and should.
- Address variants are gathered into one form — trailing slash, capitalisation, parameters.
- It is audited together with hreflang — when they conflict, the canonical wins.
A diagnostic note: if a language version never enters the index, the first place to look is the canonical declaration. A canonical pointing to the wrong place makes that page count as a copy and leaves it out entirely, giving no other symptom.