Canonicals

When Google picks a different canonical

Read the two canonical fields in URL Inspection, trace the signals Google followed and decide whether to merge URLs or make the pages distinct.

When Google picks a different canonical — nanoSEO Journal

Read both canonical fields before changing anything

Duplicate, Google chose different canonical than user is not a failed crawl. It means the URL you inspected belongs to a group of pages Google treats as duplicates, and Google indexed a different member of that group. The inspected URL can be absent from the Search index while its content is present under the selected address. This status is working as designed, not evidence that Google ignored the page. Page indexing report

Open URL Inspection in Search Console and inspect the indexed version of the failing URL. Under Page indexing, record the user-declared canonical and the Google-selected canonical. The first is the address your page or HTTP header says should represent the content. The second is the address Google chose. The live test cannot predict that choice. Google states that the canonical version can be determined only in the indexed data. URL Inspection tool

That distinction sets the job: decide whether Google's chosen URL is better, then align the signals if it is not. A request for indexing does not decide which URL should win, so settle that first.

Decide whether Google made the better choice

Treat the Google-selected canonical as a proposal, and compare it with the user-declared canonical as a visitor arriving from Search. A choice is usually sound when it picks the cleaner current URL: the preferred host, a product address without tracking parameters, the visible domain after a migration or the version that contains the full page.

When Google picks the correct URL, keep it. Update the sitemap, internal links and canonical tags to point there, and remove links that keep promoting the rejected address. Fighting a good canonical produces the contradiction that caused the problem.

Google's choice is wrong when it clusters pages with different jobs, selects an obsolete or staging host, or chooses an unrelated URL. The useful test is the page's purpose. If both addresses provide the same content and serve the same user need, one canonical is appropriate. If they answer different searches, each should stand as a distinct page with a self-referencing canonical.

The user-declared canonical does not always win. Google says it may choose a different canonical for reasons including content quality and technical signals. The tag is a strong signal, not a command. Fix canonicalisation issues

Trace the signals around the failing URL

Compare the live response and the indexed page. Start with the status code. A preferred URL that redirects, returns an error or requires a login cannot be the stable canonical target. Use curl -I to see the status and location header, then check the HTML source for the canonical element.

Look for contradictory inputs:

  • the canonical tag in the HTML source;
  • a canonical HTTP header, if the server sends one;
  • internal links pointing to the rejected variant;
  • sitemap entries for both addresses;
  • redirects between the two URLs;
  • hreflang annotations for localised versions;
  • tracking, filter, sort or session parameters;
  • page content that repeats another URL almost word for word.

Google lists redirects first, rel="canonical" second and sitemap inclusion third among the methods that influence canonicalisation. These signals can combine, which is why one correct tag may lose to three consistent signals pointing elsewhere. How to specify a canonical URL

A worked example makes the conflict easier to spot. Suppose /products/red-chair declares itself canonical. Every navigation link sends users to /products/red-chair?colour=red, the sitemap lists both URLs, and the clean URL returns a 301 to the parameter version. The tag says one thing. The links, sitemap and redirect mostly say another. Adding a second canonical tag does not repair that. Make the clean URL return 200, link to it consistently and list only it.

If multiple key templates show the same pattern, check them as a group before editing individual pages. The pre-publish page checks cover the canonical, robots and link details that belong in that verification pass.

Choose one URL when the two addresses are genuinely interchangeable. A permanent redirect is the clearest way to retire a replaced address. Google describes redirects as a strong signal that the target should become canonical. Use a 301 when the old URL should stop serving users.

A canonical tag fits when both URLs must remain live for visitors, or when changing the address would break campaigns, printed links or application routes. Keep the canonical tag on the duplicate version and include a self-referencing canonical on the preferred version. Link to the preferred URL from navigation, breadcrumbs and body copy.

Trying to settle the choice with noindex or a robots.txt block creates a different problem. Google does not recommend noindex for selecting a canonical within one site because it removes the page from Search entirely. Robots.txt can prevent Google from reading the page, and a blocked URL may still appear without its content. How to specify a canonical URL

A Google-selected URL that serves the same page correctly is not a defect. If the Google-selected URL is the better product page, article URL or hostname, the status is not a defect. Mark the chosen URL as the one to maintain, then make every other signal agree with it.

Make distinct pages look distinct

Some pages should not be merged. Two product variants may deserve separate pages because they solve different problems. Two service pages may target different locations or audiences. An article and its category page may look similar only because both use the same title pattern and boilerplate.

In that case, the user-declared canonical is only the first correction. Make the pages substantially different where they need to remain separate. Give each a specific title, H1 and opening paragraph. Add the details a visitor can use to choose between them. Use internal links with descriptive anchors that send users to the right page rather than generic links to both.

A self-referencing canonical on each distinct page is the minimum. It does not make near-identical content distinct. When the visible text is mostly the same, Google may still treat the URLs as duplicates. Google recommends ensuring that clustered pages are sufficiently different before requesting re-evaluation. Fix canonicalisation issues

Filter and sort variants usually belong in the merge group, not the distinct group. If ?sort=price and ?sort=newest return the same products in a different order, pick one indexable version and canonicalise or redirect the variants. Keep the decision tied to the user need, not the number of URLs produced by the template.

Correct contradictory signals once

Make the chosen canonical consistent everywhere before asking Google to re-evaluate anything. Declaring one URL in the HTML and another in the sitemap creates an avoidable conflict. A canonical header that disagrees with the tag is another contradiction. Internal links and hreflang annotations should point at the chosen address too.

For a migration, redirect the old URL permanently, use the new URL in the sitemap and navigation, and give the new URL a self-referencing canonical. For parameter duplicates, redirect or canonicalise to one clean address. For a wrong CMS default, fix the template or plugin instead of editing the rendered HTML by hand.

Verify the live signals after deployment. curl -I shows the response and headers. Viewing the source shows the canonical element and any rendered changes. Search Console may still show the previous crawl until Google revisits the page, so the live test is for access and indexability, not proof that the canonical has changed.

Google says re-evaluation can take up to two weeks after content fixes. Request indexing once for a small set of important URLs after the signals agree. Reserving that request for the pages that matter is more useful than sending every variant back into the queue. Fix canonicalisation issues

Check the result after the next crawl

Return to URL Inspection on the preferred URL and inspect the indexed version again. The Google-selected canonical should now match the address you chose. When the preferred URL becomes indexed and the rejected URL is marked as an alternate with a proper canonical tag, that is the expected result. The duplicate was not deleted. Google has simply attached it to the page it indexes.

Also check the live listing and internal links. Search for the preferred URL and open the result. When Google still shows the other address, or if the selected canonical changes again, look for a signal that was missed: an old redirect, a hard-coded canonical, a sitemap regenerated from the wrong source or a link template that still points to the variant.

When the page is indexed under the chosen URL but ranks worse than expected, this status is no longer the main problem. Continue with how to prioritise SEO fixes instead of reopening the canonical issue.

Leave intentional duplicates alone

Some pages should not be indexed separately. Pagination, tracking parameters, print versions, filtered category pages and near-identical regional variants often exist to serve visitors rather than to compete in Search. A canonical that points them at one useful page can be correct even when Search Console lists the alternative as not indexed.

The decision comes down to the page a person should receive. If Google's selected URL is that page, keep it, link to it and stop treating the status as an error. If it is not, change the content and signals that made Google prefer it, then wait for the next crawl.

Inspect the failing URL now and copy the Google-selected canonical into a note. Open that URL as a visitor. If it serves the same purpose, make it the one canonical target across tags, links and sitemap. If it serves a different purpose, separate the pages properly before requesting re-evaluation.

CanonicalsSearch ConsoleTechnical SEO

Published 21 November 2026 · Updated 21 November 2026. This article is part of the nanoSEO Journal.