Canonical URL

Canonical problems are plumbing problems: invisible until they leak. A site serving the same page at four URLs without a canonical splits its evidence four ways; a canonical pointing at the wrong host quietly exports authority. The fix is one declared URL per page, redirects that agree with it, and internal links that use it.

The variants multiply faster than anyone expects. Protocol (http and https), host (www and bare), trailing slash, index filenames, uppercase paths, session identifiers, and every campaign parameter a marketing team has ever appended each produce a distinct URL serving identical content. A site with four of these dimensions live at once is technically publishing sixteen addresses for every page.

Worth knowing: a canonical tag is a hint, not an instruction. Google treats it as one signal among several and will override it when the rest of the site disagrees — when internal links point elsewhere, when the sitemap lists a different URL, or when redirects contradict the declaration. This is why a canonical tag alone rarely fixes a duplication problem. The tag, the redirects, the sitemap, and the internal links all have to name the same URL, and the tag is simply the one that is easiest to check.

Auditing this site's own head section found og:url and the canonical disagreeing on the www host — years old, cosmetically invisible, and exactly the class of contradiction machines notice. The test suite now asserts they agree on every page.

For programmatic sets the stakes rise, because a canonical bug replicates across every generated page at once. That is the argument for generating canonicals from the same routing data that builds the page rather than hand-writing them per template, and for testing the assertion rather than trusting it. See indexation for the scoreboard that reveals when this has gone wrong.