One page, four addresses: duplicates and canonicals

13 August 2026 · Indexing

Most duplicate content is not plagiarism. It is one document reachable through several addresses, each created by an ordinary technical decision that nobody thought of as a decision at the time.

How one page becomes four

A single article can be served at the address with and without a trailing slash, on the bare host and the www host, over both protocols, with a tracking parameter appended by a newsletter, with a print variant, and again as the first page of its own category listing. None of that is misconduct. It is the accumulated residue of a server configuration, a link someone shared, and a plugin.

The cost is spread thin and easy to miss. Requests are spent fetching the same bytes repeatedly. Links arriving at your article are split across variants. And the machine picks one address to represent the group, which may not be the one you would have chosen.

What a canonical declaration does

A canonical link element names the address you consider authoritative for a document. It is a strong hint rather than a command: it is weighed against the other signals, and it is ignored when it contradicts them. A canonical pointing at a page with different content, or at a page that redirects elsewhere, or at a page that cannot be fetched, is discarded and you are back to guessing.

Rules that keep canonicals believable:

• Every page names an address, including the one it names for itself. A self-referential canonical on the preferred address is the cheapest way to remove ambiguity.
• The named address returns a 200 status and the same content.
• The named address is the one in your internal links and your sitemap. Three different opinions in three places is worse than no opinion at all.
• Never point a whole section at the homepage. That is a request to have the section treated as noise.

When to redirect instead

A canonical is for cases where several addresses must keep working. When only one should exist, a permanent redirect is the honest tool: protocol and host normalisation, a trailing-slash policy, a moved article. Redirect once, to the final address, without a chain through two intermediates.

Pagination and near-duplicates

Pages two, three and four of a listing are not duplicates of page one; they are different documents with the same shape, and they should say so with their own titles. Do not point them at page one with a canonical, and do not build a view-all page you then contradict. Where near-duplicates exist for a reason, such as regional variants, the differences should be visible in the copy, not only in a tag.

Permalink: /notes/duplicate-urls-and-canonicals/