Canonical URL

The canonical link relation was described by the IETF in RFC 6596, in 2012: a page can name the preferred version of itself among resources with duplicate content^[1]. Search engines read it as one of several signals.

Duplicates are normal: tracking parameters, filters, sort orders, http and https, closing slashes. Google lists the usual causes as region and device variants, protocol variants, site functions such as sorting and filtering, and accidental variants^[2]. Without a clear canonical, ranking signals are split across the copies, crawling is wasted, and the URL in the results may be the wrong one.

A declared canonical is a hint, not a command. Google lists the ways of stating one in order of strength: a redirect and a rel="canonical" link are strong signals, and listing a URL in a sitemap is a weak one^[3]. Google reports both the canonical you declared and the one it selected, and they differ when the signals disagree.

The same discipline applies to host names. In QueryRev’s study of the top sites, 319 domains could be read only on their www. host, after the bare domain failed to answer^[4]: two addresses for one site, which every crawler has to resolve before it reads anything.

On a shop with filtered navigation, thousands of parameter URLs can stand behind a few hundred real pages. A canonical on each variant tells search engines which page they all stand for, so crawling and ranking signals go to the page that should rank.

Crawl the site and record each page’s canonical target, then check it: does the target answer 200, is it indexable, does it point to itself? In Search Console’s URL Inspection, compare the user-declared canonical with the Google-selected one. Count the pages canonicalised elsewhere, canonical chains, and canonicals that point to a redirect or an error.

Then check that every signal agrees: the canonical link, the redirects, the internal links and the sitemap should all name the same URL. Google advises against naming different canonicals for one page through different methods^[3].

Read the canonical as the search engine reads it: from the HTML the server sends, and from the HTTP header where one is set. A canonical added by a script after the page loads may be seen late or not at all, and two different canonicals in the HTML and the header cancel each other out.

Sample data: an invented site, to show the method.

A collection page can be reached at /collections/bed-frames, with ?sort=price and with ?utm_source=mail. All three declare the first as canonical. One page is indexed; the two variants pass their signals to it.

Internal links and the sitemap list only the clean URL, so the declared canonical and every other signal agree.

When a sort option is added later, its URLs take the same canonical from the collection template, so the rule holds without a change on each page.