Canonical Tag SEO: Why Google Ignores Yours and How to Fix It

Here is the sentence that explains most canonical tag problems, and it comes from Google’s own canonicalization documentation: “indicating a canonical preference is a hint, not a rule.” Your rel="canonical" does not decide anything. It is one input into a selection Google makes on its own, and when the rest of your signals point somewhere else, Google follows the rest of your signals.

That single distinction separates the sites where canonicalisation works from the ones where it quietly does not. Below is what the documentation actually says, the order Google ranks the competing signals in, and the conflict matrix I run when a client’s canonical is being ignored.

What a canonical URL is, in Google’s words

Google defines it plainly: “A canonical URL is the URL of a page that Google chose as the most representative from a set of duplicate pages.” Note the verb. Google chose. Not you.

Two further statements from the same documentation matter for anyone auditing a site. First, duplication is not in itself a violation: some duplicate content is normal and is not treated as spam. Second, there is a crawl consequence. Google says the canonical page “will be crawled most regularly; duplicates are crawled less frequently in order to reduce the crawling load on sites.” A page Google has clustered as a duplicate is not just missing from results, it is being visited less often, which is why canonical problems on large sites compound rather than sit still.

The signal hierarchy: what actually outranks what

Google’s consolidation documentation lists the methods and grades them. This is the part most canonical advice leaves out, and it is the part that tells you what to reach for.

Method	Strength, per Google	What it is good for
Redirect (301 or 308)	“A strong signal that the target of the redirect should become canonical”	Retiring a URL permanently. The only method that also removes the duplicate from circulation.
rel=”canonical” annotation	“A strong signal that the specified URL should become canonical”	Keeping both URLs reachable by users while consolidating indexing.
Sitemap inclusion	“A weak signal that helps the URLs that are included in a sitemap become canonical”	Supporting a decision already made elsewhere. Never sufficient on its own.
robots.txt disallow	Not a canonicalisation method at all	Nothing here. Google explicitly warns against using it for this purpose.
noindex	Not a canonicalisation method	Removing a page from results. It does not consolidate anything into another URL.

Redirects and canonical annotations sit at the same stated strength. What separates them in practice is that a redirect removes the alternative, and a canonical annotation leaves it standing and competing. Every other signal on a page keeps voting after you add the tag.

The conflict matrix: why your canonical is being overruled

When a canonical is ignored, it is almost never because the tag is malformed. It is because something else on the site is making a louder claim. This is the matrix I work through, in order, on a canonical audit. Each row is a real conflict I have found on live sites, and the resolution column is what Google’s documented signal hierarchy predicts will win.

Your canonical says	The conflicting signal	Likely winner	The fix
Page A is canonical	Internal links across the site all point at page B	Page B	Change the internal links. They are the site telling Google which page it treats as real.
Page A is canonical	The XML sitemap lists page B	Page A, but weakly	Remove B from the sitemap. A weak signal contradicting a strong one still muddies the cluster.
Page A is canonical	Page A 301 redirects to page B	Page B	The redirect wins. Decide which behaviour you want and remove the other.
Page A is canonical	HTTP version canonicalises to HTTPS elsewhere	The HTTPS URL	HTTPS is a documented selection factor. Make every canonical absolute and HTTPS.
Page A is canonical	The HTML head and the HTTP header give different canonicals	Unpredictable	Google warns against specifying different canonicals through different methods. Pick one method.
Page A is canonical	Pages A and B are not actually duplicates	Neither, the tag is ignored	Canonicalisation clusters duplicates. Two genuinely different pages will not cluster no matter what you annotate.
Page A is canonical	Page A carries noindex	Nothing is indexed	Google lists this as a mistake to avoid. Remove one of the two.
Self-referencing canonical	The URL is blocked in robots.txt	Google never sees the tag	A blocked page cannot be read, so its canonical cannot be honoured. Unblock or redirect instead.

The pattern across all eight rows is the same. A canonical tag is a claim about a relationship, and the rest of the site is making claims too. Auditing a canonical in isolation tells you almost nothing.

The mistakes Google names directly

Google’s documentation lists specific errors, and they are worth quoting rather than paraphrasing because each one shows up in audits constantly:

Do not use robots.txt or a URL removal tool to try to force canonicalisation. Neither consolidates signals.
Do not specify different canonical URLs for the same page using different methods, for example one value in the head and another in the HTTP header.
Do not specify a URL fragment as the canonical. Anything after a hash is not a separate URL to Google.
Do not use noindex as a way of steering canonical selection.

A worked example: a faceted category with four live variants

A common real set. One category, four reachable URLs, all serving near-identical content:

https://example.com/boots/
https://example.com/boots/?sort=price
https://example.com/boots/?colour=black
https://example.com/boots/?sort=price&colour=black

The configuration that holds up, and the reasoning for each line:

<!-- on all four URLs -->
<link rel="canonical" href="https://example.com/boots/">

Absolute, HTTPS, no fragment. A relative canonical resolves against the current URL and will silently self-reference on the wrong page.
The clean URL also carries a self-referencing canonical. It costs nothing and it removes ambiguity if the page is ever reached with a tracking parameter appended.
Only the clean URL goes in the sitemap. The weak signal now agrees with the strong one instead of contradicting it.
Internal links point at the clean URL. This is the row in the matrix above that overrules more canonicals than any other.
Nothing is blocked in robots.txt. A blocked variant is a variant whose canonical Google cannot read.
Nothing carries noindex. Mixing noindex with a cross-page canonical is on Google’s own list of mistakes.

Note what is not in that list: a faceted variant that genuinely produces a different product set is not a duplicate, and canonicalising it away deletes a page that could have ranked. Cluster what is duplicate. Leave what is not.

How to find out which URL Google actually chose

You do not have to guess. Search Console’s URL Inspection tool reports two separate fields: the user-declared canonical, which is your tag, and the Google-selected canonical, which is the decision. When those two disagree, you have your answer and the matrix above tells you where to look.

The Page Indexing report gives you the same thing at scale. “Duplicate, Google chose different canonical than user” is the category that names the problem outright. “Alternate page with proper canonical tag” is the state you want, and it means the tag is being honoured.

When a canonical problem looks like a penalty

This is why the topic sits inside a recovery cluster rather than a tidy technical corner. A site that loses a large share of its indexed pages to an unintended canonical cluster sees exactly what an algorithmic demotion looks like from the outside: traffic down, rankings gone, nothing in Search Console under manual actions. The difference is that this version is reversible in days rather than months, and the diagnosis takes one look at the Page Indexing report.

Before anyone starts auditing links or drafting a reconsideration request, the indexing state has to be ruled out. It is the cheapest check on the list and it is the one most often skipped.

Canonical Tag SEO: Why Google Ignores Yours, and What to Do About It