Canonical tags and duplicate content: how Google chooses a URL, and what to do when it chooses a different one

A canonical tag names the URL you want indexed when several addresses serve the same content. Google treats it as a strong hint rather than a command. It combines your tag with your redirects, your internal links, which URL is in your sitemap, which version is on HTTPS, and which one looks cleaner, then picks a single Google-selected canonical to index and show. Duplicate content inside your own site is therefore a selection problem, not a penalty: Google indexes one URL, folds the signals from the rest into it, and the others stop appearing. Your job is to make every signal point at the same URL so the choice is the one you wanted.

This page covers how Google makes that choice, how to steer it, and the fix for each duplicate pattern. For the tag syntax in full and the mistakes that cost pages their index spot, see canonical tags, what canonical means in SEO and how to set one.

User-declared and Google-selected: two different things

Search Console reports both, and the gap between them is the whole diagnosis.

Term	What it means	Where you see it
User-declared canonical	The URL your page asks for in its rel=canonical tag	URL Inspection, under Page indexing
Google-selected canonical	The URL Google actually indexed for that content	URL Inspection, same panel

When the two match, the page is settled. When they differ, Search Console labels the duplicate "Duplicate, Google chose different canonical than user", and that label is telling you something useful: another signal on your site is louder than your tag. Find it and the tag starts winning. Google documents the whole mechanism in its guide to consolidating duplicate URLs and in its reference on canonicalization.

The signals Google weighs alongside your tag

• Redirects. A 301 is the loudest signal there is. It beats a tag every time, because a redirect removes the alternative rather than ranking it lower.
• Internal links. If three hundred links across your site point at the parameter version and twelve point at the clean one, the tag is arguing against your own navigation.
• The sitemap. Only canonical URLs belong in it. A sitemap that lists both versions is a vote for both.
• HTTPS over HTTP, and a shorter, cleaner URL over a longer one with tracking on the end.
• Content similarity. If the two pages are not really the same page, Google keeps both and your tag is dropped. That is correct behaviour, and it is the most common reason a tag is ignored.

The markup, in the three places it goes

In the head of an HTML page, absolute URL, once:

<head>
  <link rel="canonical" href="https://example.com/boots/leather-chelsea" />
</head>

For a file with no head, such as a PDF, in the HTTP response header:

HTTP/1.1 200 OK
Content-Type: application/pdf
Link: <https://example.com/guides/boot-sizing.pdf>; rel="canonical"

Verify the header landed before you move on:

curl -sI https://example.com/guides/boot-sizing.pdf | grep -i '^link:'

And on every indexable page, a self-referencing canonical that names the page itself. It costs nothing and it settles the question of parameters, uppercase variants and trailing slashes before anybody asks it.

Canonical, redirect or noindex

Three tools, three jobs. Picking the wrong one is how pages disappear.

Situation	Use	Why
Both URLs must stay reachable and serve the same content	rel=canonical	Consolidates indexing, keeps both usable
One URL has permanently replaced another	301 redirect	Removes the duplicate outright, strongest consolidation
A page should never appear in search at all	noindex	Removal, not consolidation
Thin filter pages you do not want crawled heavily	Canonical plus fewer internal links	Keeps crawling proportionate to value

Never combine a canonical and a noindex on the same URL. The two instructions contradict each other, and the result is whichever one Google reads as dominant on the day.

The duplicate patterns, and the fix for each

• Tracking parameters. Self-referencing canonical on the clean URL, and the clean URL in every internal link and in the sitemap.
• HTTP, HTTPS, www and non-www. Pick one host and redirect the other three with a 301. This is a server job, not a tag job.
• Paginated listings. Each page canonicals to itself. Pointing page two at page one hides products that only appear on page two.
• Faceted navigation. Canonical the useful combinations to themselves and the long tail of filter permutations back to the category. Combine that with internal linking discipline, which crawl budget optimization covers in detail.
• Product variants. One canonical product URL with variants selected on the page, or separate URLs when a shopper genuinely searches for the variant. Shopify SEO walks through the variant URL decision on a platform where it bites hardest.
• Syndicated articles. Ask the republishing site for a cross-domain canonical pointing at your original. Where they cannot set one, a prominent link to your version is the next best arrangement.
• Staging and development copies. Password protection, not tags. A staging site that canonicals to production is still crawlable and still leaks.

Diagnosing a canonical problem in Search Console

1. Open URL Inspection, paste the URL you want indexed, and read the user-declared and Google-selected lines.
2. If they differ, open the Google-selected URL and look at what is pointing to it: internal links, the sitemap, any redirect.
3. Check the Page indexing report for the two labels that matter: "Duplicate, Google chose different canonical than user" and "Alternate page with proper canonical tag". The second one is a success message, not an error.
4. Fix the louder signal, then request indexing on the URL you want kept.
5. Re-inspect in a week. Canonical consolidation settles over a crawl cycle, so check again rather than changing the tag twice.

canonical tags and duplicate content: what readers write in to ask

Does duplicate content get a site penalised?

Ordinary duplication inside a site is handled by selection: Google picks one URL and consolidates the rest. Penalties belong to deliberate schemes, such as scraping other people's pages at scale, which Google addresses in its spam policies.

Why is Google ignoring my canonical tag?

Almost always one of four reasons: a redirect or internal links point elsewhere, the two pages are not similar enough to be duplicates, the tag uses a relative URL, or the tag points at a URL that is itself redirected or blocked. Check in that order.

Should every page have a self-referencing canonical?

Yes, on every page you want indexed. It is the cheapest insurance against parameters and case variants creating duplicates you never planned for.

Can I canonical to a page on another domain?

Yes. Cross-domain canonicals are the documented way to handle syndication, and they work the same way: a strong hint that the other domain holds the master version.

How long does a canonical change take to show?

Google has to recrawl both URLs and reassess them, so work in crawl cycles rather than days. Requesting indexing on the preferred URL shortens the wait.

Does a canonical pass link signals?

Consolidation is the point of it. Signals from the duplicates are attributed to the chosen URL, which is why a clean canonical set usually ranks better than the same content spread across six addresses.

Which URL does the answer come from?

Canonicals tell Google which version of a page to show. An assistant asked the same question has its own way of choosing between your versions, and the one it cites may be a parameter URL, a syndicated copy or the page you chose. Open the sources in an answer and read the exact address. If it is the wrong one, your signals are less clear than Search Console suggests. AI Knows Us lists the pages each assistant relied on, address by address.