Why does Google sometimes change its mind about your canonical URL?

Does Google fix the canonical after the first evaluation?

No. Google never makes a definitive decision on the canonical URL of a group of duplicate pages. Martin Splitt clarifies: algorithms continually reevaluate signals with each crawl to detect any changes in content, structure, or external signals.

This ongoing reevaluation means that a URL chosen as canonical today may be replaced tomorrow by another version, even without manual intervention on your part. The system is dynamic, not static.

What causes these switches between URLs?

The issue arises when two versions present very close duplication scores — for example, 0.49 versus 0.51. In this gray area, Google lacks a clear enough signal to consistently make a decision.

Two main triggering factors: crawl order (which URL Googlebot visits first during a given session) and content modifications, even minor ones, can slightly invert the scores. If you add a phrase or change a title on version A, it may suddenly switch ahead of version B on the next crawl.

How does this instability pose an SEO problem?

This fluctuation generates unpredictable indexing variations. You may notice that Google sometimes indexes the URL with parameters and other times the clean version, without any apparent logic. Rank tracking tools may show artificial fluctuations if the indexed URL changes.

Worse, ranking signals (backlinks, UX signals, authority) get fragmented across multiple URLs instead of concentrating on just one. You lose consistency in accumulating positive signals, diluting your ranking potential.

• The canonical is never permanently fixed — Google reevaluates with every crawl
• Close duplication scores (e.g., 0.49 vs 0.51) create a zone of instability where the choice easily shifts
• Crawl order and content modifications directly influence which URL will be retained
• This instability can fragment your ranking signals across multiple versions and hinder your performance
• Without strong and consistent signals (canonical tag, redirects, internal linking), you leave Google to arbitrate on its own — with variable results

Is this statement consistent with real-world observations?

Yes, completely. Canonical URL fluctuations are regularly observed in Search Console, especially on e-commerce or media sites that generate many pages with minor variations (filters, pagination, UTM parameters).

What's new here is the official confirmation that Google never makes a definitive decision. Previously, many SEOs believed that once the canonical stabilized after a few weeks, the matter was settled. False — the system is constantly reevaluating, and if your signals remain ambiguous, the choice can switch with every crawl.

What nuances should be added to this statement?

Splitt talks about “very close” scores (0.49 vs 0.51), but Google provides no actual numbers or public thresholds. We don’t know how similar two pages must be to enter this gray area. [To be verified] : does 85% similarity suffice? 95%? No numerical data is provided.

Another point: the statement implies that explicit signals (canonical tag, 301 redirects) weigh less than expected when the contents are nearly identical. If you have a clean rel=canonical but both pages remain crawlable and their content slightly varies, Google may ignore your directive and choose the other version. This is not explicitly stated, but it’s the practical implication.

In what cases does this rule not apply?

If your canonicalization signals are strong and converging, the problem disappears. Specifically: a strict 301 redirect, a clear canonical tag, consistent internal linking always pointing to the same version, and XML sitemaps that only list one URL.

In this case, even if two pages have similar content, Google has no reason to switch — the technical signals are clear. Fluctuation occurs mainly when you leave multiple versions crawlable and indexable without unambiguous directive. If you properly block duplicates (robots.txt, noindex, or redirect), the problem no longer arises.

What specific actions should you take to avoid these switches?

First action: audit all your duplicate or nearly duplicate pages. Identify groups of URLs with similar content (filters, pagination, parameters, mobile/desktop versions if they differ slightly). Use Search Console to find pages marked as “Detected, currently not indexed” or “Excluded by the canonical tag” — these are often signals of fluctuation.

Next, strengthen your canonicalization signals. Add or verify the rel=canonical on each duplicated page, pointing to the version you want to index. If possible, 301 redirect unnecessary versions instead of leaving them crawlable. Clean your XML sitemap to only list canonical URLs — never the variants.

What mistakes should you absolutely avoid?

Do not leave multiple versions of the same page crawlable AND indexable without a clear directive. This is the classic mistake: you think Google will “understand on its own” which version to prioritize, but if duplication scores are close, it will hesitate and switch regularly.

Avoid cascading or contradictory canonical tags. For example, page A points to B as canonical, but B points to C. Or worse, A and B point to each other. Google ignores these inconsistent signals and chooses on its own, with the risks of fluctuation that this implies.

How can you confirm that your site is well stabilized?

Monitor Search Console, under the “Coverage” and “Pages” tabs. Look for URLs that regularly switch from “Indexed” to “Excluded by canonical” or vice versa. If you see frequent indexing variations on the same group of pages, it’s a symptom of fluctuation.

Also, use a crawler (Screaming Frog, Oncrawl, Botify) to check that your canonical tags are all pointing to the same version and that no chain redirects exist. Test URL inspection in Search Console on your critical pages: Google should consistently return the same canonical URL with each test, not variable results.

• Audit all groups of duplicate or nearly duplicate pages
• Add or verify the rel=canonical on each duplicated page, pointing to the unique version to be indexed
• 301 redirect unnecessary versions instead of leaving them crawlable
• Clean the XML sitemap to only list canonical URLs
• Check for the absence of cascading or contradictory canonical tags
• Monitor Search Console for frequent indexing variations on the same URLs