Greta Lindholm

A site should nominate one address for each piece of content and make every other route lead there, with the two mechanisms telling the same answer. A redirect chain, though, is weaker than a single hop, and a chain that eventually arrives somewhere is often treated as a soft signal rather than a firm one. The tag is a hint declared in the page. It is ignored when it points to a redirect, because the tag would then be describing a page that is not the one being served. It is also self-referential in a way that confuses people: a canonical on a page pointing to itself says nothing, and the useful case is the page pointing to the preferred version. Duplicate detection has no single answer, which is why tools disagree. A page reachable at three addresses is a duplicate of all three, including itself. Parameter handling, trailing slashes, case in the path and the protocol all create addresses that serve identical content and that crawlers treat as distinct unless told otherwise. Solving this by changing a slug is the one thing I refuse to advise. It discards the accumulated history of an address, and the address is often already ranked. My rule is that one address, one page, one nominated canonical, and every other route reaching it by redirect. Everything else on this subject is a variation on that or a deviation from it. Detection tools disagree because they sample differently, which is why the canonical decision should be made from a crawl of the site rather than from any one tool's output.