Canonical tag

Usually without anyone intending it. A tracking parameter appended to a shared link. A session ID. A shop filter that produces ?colour=blue. A printer-friendly version. The same product sitting in two categories. A trailing slash, or its absence.

To a search engine each of those is a separate URL that happens to serve identical content — and it has to decide which one to index. The canonical tag is how you answer that question instead of leaving it to be guessed.

Google treats rel=canonical as one signal among several, not as a command. Internal links, the sitemap, redirects and the tag itself are weighed together — and if they disagree, Google picks its own canonical and reports the difference in Search Console as Alternate page with proper canonical tag or, less happily, Duplicate, Google chose different canonical than user.

The second message is nearly always a consistency problem: the page declares one canonical while every internal link and the sitemap point somewhere else. The tag alone does not outvote your own linking.

• Self-reference every page. A page whose canonical points at itself is the normal case, not a special one. Rank Math and Yoast do this automatically.
• Use absolute URLs. Relative canonicals are permitted and reliably misconfigured.
• Link the canonical, always. Every internal link, every sitemap entry, the same address the tag names.
• Do not chain them. A canonical pointing at a page whose canonical points somewhere else is ignored.
• Do not canonicalise paginated pages to page one. Page two is not a duplicate of page one; each should reference itself.
• Never combine it with noindex on the same URL. The two say contradictory things.

The distinction is whether visitors still need the other URL. If nobody should land on the old address any more, use a 301 redirect — it is stronger and unambiguous. If both addresses have to keep working, as with a filtered shop listing or a print view, the canonical is the right tool.