www versus bare domain and canonical URLs in 2026: a 600-host measurement

On 27 August 2026 we downloaded the home pages of 600 hosts and checked for <link rel="canonical">:

A missing tag is not in itself an error: for a home page a search engine usually picks the right address anyway. The error appears where the same page is reachable at several addresses — with and without www, over HTTP and HTTPS, with and without a trailing slash, with campaign parameters — and nothing says which one is primary.

Caveats: these are domains somebody brought to our tools, not a random slice of a zone; comparing zones is sound as they were collected identically. Only the home page was checked — inner pages may differ.

Among those that did declare a canonical:

The difference is visible: in .ru nine in ten point at the bare domain, in .com only two thirds. Site age is the likeliest explanation — www is a holdover from when the subdomain denoted the web service specifically, and .com holds historically more of those domains.

In practice the choice between www and bare does not affect ranking. Only one thing matters: that a choice was made, held consistently, and backed by a redirect. Technical differences exist (a bare domain cannot carry a CNAME at the apex, and its cookies reach every subdomain), but they have nothing to do with search.

The canonical points at a different host. Found on 7 pages in .ru and 6 in .com. Usually this is the trace of a template copied from another project, or an under-configured mirror. The consequence is direct: the page tells a search engine its real version belongs to someone else, and asks not to be indexed.

The canonical is declared over HTTP while the page is served over HTTPS. Three cases in .ru, one in .com. This is the residue of an HTTPS migration that never reached the template. A search engine will most likely ignore the instruction, but relying on its forbearance is unwise.

What unites both errors is that they are invisible to the eye: the page opens, looks right, nothing breaks. They can only be found by inspecting the markup — which is exactly why they survive for years.

1. Pick one primary form of the address — with or without www — and hold to it everywhere: in the canonical, the sitemap, internal links and analytics.
2. Add a permanent redirect (301) from every other variant to the chosen one. To confirm the chain is short and lands where intended, use the redirect checker.
3. Use an absolute URL in the canonical — scheme and host included. A relative tag is formally permitted but more often results from a template mistake than a decision.
4. Check the tag does not point at someone else’s domain — especially if the template came from another project. It is the first thing to look at when pages drop out of the index inexplicably.
5. Make sure the canonical scheme is HTTPS if the site runs on HTTPS.

To see what your page carries, use the page audit.