When Google chooses a different canonical from the one you declared, what is happening? You told search engines which version of a page should count as the main one, and Google decided a different URL was a better representative. Your rankings, links, and reporting then attach to a page you did not intend to promote.
Who meets this problem? Anyone whose site has similar pages, multiple addresses for the same content, or a long history of redesigns. Why does Google override you? Because a canonical tag is a strong hint, not an order, and when your other signals disagree with the hint, Google follows the signals.
Where should you start? Not with rewriting tags at random. Start by identifying which cause is at work on your site, because each cause has a specific fix, and applying the wrong fix usually makes the disagreement worse.
The clearest evidence sits in Search Console, in the page indexing report and the URL inspection tool, where Google lists the user-declared canonical next to the Google-selected canonical. When those two differ, this is your problem, whatever your tags say.
The symptoms elsewhere are confusing by nature. Traffic lands on an odd URL variant, a secure page loses ground to an insecure twin, or a product appears under a filtered address. Reports seem inconsistent because the credit for the page keeps moving between addresses.
Google compares your declared canonical against everything else it can see: internal links, sitemap entries, redirects, and the content itself. If most of those point one way and your tag points another, Google trusts the weight of evidence. Fixing the problem therefore means aligning evidence, not arguing louder with a tag.
Cause: Two pages are so similar that Google sees no reason to honour your choice between them. Printer versions, location pages with swapped city names, and product pages differing only by colour are classic examples. If the pages are near-identical, Google picks whichever it found first or trusts most.
Fix: Make the intended canonical clearly the strongest version. Add unique, useful content to it, merge the weaker twin into it where possible, and redirect the twin if it has no independent purpose. A canonical tag between two almost identical pages is a weak tool; differentiation is the real remedy.
Cause: Sort orders, filters, session identifiers, and tracking parameters create many addresses for one page. Some parameter version picks up links or gets listed in a sitemap, and Google concludes that version is the one the site really uses.
Fix: Keep parameter URLs out of sitemaps, link internally only to the clean version, and use canonical tags consistently from every parameter variant to the clean URL. Check which variants have actually been seen by using our Google Cache Checker alongside Search Console, so you are fixing real indexed variants rather than theoretical ones.
Cause: The same page answers at http and https, with and without www, and sometimes on a staging or regional host. If redirects are incomplete, or the insecure version carries more old links, Google may select the variant you thought you had retired.
Fix: Choose one host and protocol, redirect every other variant to it permanently, and make canonical tags self-referencing on the chosen version. Then update internal links so nothing on your own site still points at the retired variants. Mixed signals from your own pages are the most common reason this cause persists.
Cause: The declared canonical contradicts the rest of the site. The sitemap lists one URL, internal links favour another, a redirect chain lands somewhere else, and syndicated copies carry no canonical back to you at all. Faced with that confusion, Google picks the version the evidence supports, which may not be yours.
Fix: Audit the signals as a set and make them unanimous. One URL in the sitemap, one URL in internal links, one URL in the canonical tag, and redirects that point straight at it. Where partners republish your content, ask for a canonical or link back to your original, and keep your own house in order first, because practices that mislead engines about a page's identity sit close to the behaviours covered by Google's spam policies, even when the confusion is accidental.
Further worked examples of signal audits, and guides to related indexing problems, are collected on the Seovaro blog for when you want to go deeper on a particular cause.
Before your next fix, try the Google Cache Checker linked above on your problem URLs: confirm which variants Google actually holds, match each one to a cause in this catalogue, and apply the paired fix instead of changing tags and hoping.
No. It is a strong hint. Google weighs it against internal links, sitemaps, redirects, and content similarity, and it will choose a different canonical when those signals consistently point elsewhere.
Google must recrawl the affected pages and their signals first. On regularly crawled sites that can be days, while deep or rarely visited pages may take several weeks to reflect the change.
For most sites, yes. A self-referencing canonical on the preferred version removes ambiguity, especially where parameters or shared links might otherwise create competing variants of the same page.
Check for syndication without a canonical back to you, scraped copies, and near-identical content elsewhere. Strengthen your original with unique material and internal links, and ask republishing partners to credit your URL.