The Canonical Tag: A Request Google Can Refuse

On September 14, 2026, I ran a simple search on Google. I asked for the pages in the “about us” section of seogenome.com, our company site in Brazil. Three came back. ¹

Two of them carry a request to Google in their code. The request says: show another address instead of this one. One points to danielsocrates.com.br, the other to pollymelo.com.br. ^{1 2}

Google showed both anyway, under the seogenome.com address. ¹

It did what its own documentation warns it may do. ³ The canonical tag is a request, and Google makes the call.

This piece shows how that request works. What Google reads before deciding, how much each signal weighs, and which requests it tends to turn down. And where you check, in Search Console, whether yours was granted.

One boundary before we start. Type “canonical” into Google in Brazil, where we are based, and you land somewhere else. Every result on the first page was about Canonical, the British company that publishes Ubuntu. ⁴ Here the subject is the SEO canonical: pointing to the canonical URL of a page.

What a canonical URL is

Picture a newsroom that receives the same photo in five envelopes. One is cropped, another is in black and white, another carries the photographer’s name. The editor will run only one. You can send a note saying which one you prefer. The editor reads the note, looks at all five, and chooses.

In search, the editor is Google, and the envelopes are addresses. The same content often answers at more than one URL. Over HTTP and HTTPS, in the mobile and the desktop version, with filters and with sorting. ³

Google groups the pages that seem to be the same, or whose primary content is very similar. Then it picks the one that is most complete and useful for searchers. ³ The chosen one is called the canonical URL. The whole process is called canonicalization.

The choice has consequences. Google uses the canonical page as the main source to evaluate content and quality. ³ It gets crawled most regularly, and the duplicates less often. ³ What that crawling costs the crawler is in the piece on crawl budget.

The note is your preference. The documentation says it in one line: indicating a canonical preference is “a hint, not a rule.” ³ Google may choose a different page, for various reasons. ³

This idea of a request comes from the specification itself. RFC 6596, published by the IETF in April 2012, defines the canonical link relation. Informally, it describes it as the author’s preferred version. ⁵ Google states that it follows that specification. ⁶

A stronger idea circulates in the industry. Well-ranked texts on Google in Brazil say the tag guarantees which version gets indexed, or that it keeps the duplicates out of the index. ⁴ Google’s documentation describes a request that carries a lot of weight, with no guarantee.

How Google chooses: the signals and the weight of each

Besides the note, the editor looks at what each envelope says and at which photo everyone points to. Google does the same with a handful of factors.

The documentation lists four. Whether the page is served over HTTP or HTTPS, redirects, presence in the sitemap, and the canonical tag. ³ The page on how to specify a canonical gives the weight of the three methods that depend on you, in order of strength: ⁶

1. Redirect. A strong signal that the target should become canonical.
2. Canonical tag. A strong signal that the specified URL should become canonical.
3. Sitemap. A weak signal.

The three stack. Using two or more raises the chance that your preferred URL shows up in search. ⁶

There are also signals that come from how the site is built. Google prefers HTTPS over HTTP by default. ⁶ It drops that preference in some cases. An invalid certificate, an HTTPS page that redirects to HTTP, or a canonical pointing to the HTTP version. ⁶ And it prefers URLs that are part of a reciprocal hreflang cluster. ⁶

Internal links come in as good practice. Google asks you to always link to the canonical URL. Linking to the same address consistently helps Google understand your preference. ⁶

One detail takes weight off many site owners’ shoulders. None of these methods is required. With no preference stated, Google works out the best version to show on its own. ⁶

Stating one pays off in four situations, by Google’s own list. ⁶ When you want to choose the address that shows up in search. When you want to consolidate the links that reach the duplicates into a single URL. When you want to measure the content in one place. And when you want to spare the crawler from crawling duplicates.

The ways to ask, and what each one costs

The documentation compares the methods side by side, with pros and cons. ⁶

The tag in the code. It is a <link> element inside the page’s <head>, pointing to the canonical. The example below uses a made-up address:

<link rel="canonical" href="https://www.example.com/dresses/green-dress/" />

It maps as many duplicates as you want. The cost is maintaining that map on a large site, or on a site whose URLs change often. And it only works on HTML pages. ⁶

The HTTP header. The server sends the preference in the response itself. It does not add to the page size, and it works for files like PDFs. ⁶ The maintenance cost is the same as the tag’s. ⁶

Between the tag and the header, Google asks you to pick just one. Both together work, and they also leave room to point to one URL in one and to another URL in the other. ⁶

The sitemap. Every page listed in the sitemap is suggested as canonical. ⁶ It is the simplest method to maintain on a large site. The cost is strength: the signal is weaker, and Google still has to discover the duplicates of each canonical. ⁶

The redirect. It tops the strength list. Google recommends this method only for retiring the duplicate. ⁶ If the duplicate needs to stay live, the redirect is off the table.

Three cautions apply to the tag. ⁶ First, an absolute URL, with https and the domain, and never just the path. Second, the tag is only accepted inside the <head>, so that part of the HTML has to be valid. Third, on a site built with JavaScript, the canonical should come in the source HTML, with no script changing it later.

If you use WordPress, Wix, or Blogger, you may not be able to edit the HTML directly. In those cases, the canonical usually comes from a setting in the platform itself. ⁶

The requests Google tends to turn down

Back to the editor. The editor honors the note when all five envelopes really hold the same photo. When the note points to a different photo, there is no reason to honor it.

A canonical pointing to a different page. The most important rule is in the Search Console help. If the declared canonical is not similar to the current page, Google will never choose it. A duplicate has to be similar to the canonical. ¹¹ The RFC puts it another way. The target must hold the same content as the page, or a superset of it. ⁵

Google’s blog recorded an example in 2013. A category page pointed its canonical to the day’s featured article. If Google accepted it, the category would vanish from search. ⁹ The fix offered was for the category to point to itself, or to nothing at all. ⁹

A canonical pointing to a URL with an error or a noindex. The same post asks for two checks on the target. It has to exist, with no error and no soft 404. And it cannot carry a noindex instruction. ⁹ The post warns that parts of it may be out of date. In my reading, both checks still hold, because the canonical is the page that goes into the results. ³

The neighboring mistake runs the other way. Some people use noindex to keep a duplicate from being chosen as canonical. Google does not recommend it. Within the same site, noindex takes the page out of search entirely, and the canonical tag is the preferred solution. ⁶

A canonical from page 2 to page 1. In a paginated listing, page 2 shows other items from the list. The pagination guide is direct: do not use the first page as the canonical for the sequence, and give each page its own canonical. ⁸ The RFC names the risk: loss of data, because only the content of the first page would be processed. ⁵

Two canonicals on the same page. It happens when the theme and the SEO plugin each write their own. In the 2013 post, Google says it will likely ignore all of them. ⁹ A tag written in the body of the page, outside the <head>, does not count either. ^{6 9}

The same request down two paths. The sitemap points to one URL and the tag to another, for the same page. Google says not to do this. ⁶ The same list of don’ts includes robots.txt, the URL removal tool, and the fragment after the # sign. ⁶

Two cases already have their own piece. Search filters and the same listing repeated across listing sites are in SEO for real estate agencies. That piece explains why the canonical stays as a backup for filters. And why Google does not recommend the tag for content syndicated to partners. ⁷

A specialty page and a practitioner page with the same core text are in SEO for clinics. The symptom the industry calls cannibalization shows up, in the documentation, as a cluster of similar pages. ³

Where to see Google’s choice

The note sits in your code. The editor’s answer sits in Search Console, in two places.

In URL Inspection. Paste the address and open the page indexing section. Two fields tell the story. ¹⁰ User-declared canonical shows what your site asked for. Google-selected canonical shows what Google chose. ¹⁰

If the two match, the request was granted. If the selected one is the inspected page itself, Google thinks that page does not resemble any other. ¹¹

Two warnings from the help page save time. The live test cannot predict the canonical choice, which only shows up in the indexed data. ¹⁰ And that value can be a few hours out of date. ¹⁰

In the Page indexing report. Three reasons deal with canonicals. ¹¹

1. “Alternate page with proper canonical tag.” The page points to the right canonical, which is indexed. Nothing to do.
2. “Duplicate without user-selected canonical.” There was no request, and Google chose another page as canonical. The help says this is not an error.
3. “Duplicate, Google chose different canonical than user.” There was a request, and Google preferred another URL.

In the third case, the help page says to open three pages in your browser. The current one, the declared one, and the selected one. ¹¹ If the declared one does not resemble the current page, the request is wrong.

If the pages are similar and Google chose another one, the canonicalization troubleshooting page suggests looking at content quality and technical signals. ⁷ After a fix, Google may hold the pages in a duplicate cluster for up to two weeks. ⁷ Requesting indexing asks for a new evaluation, and it has a quota, so save it for your most important URLs. ⁷

One caution for anyone measuring. If the chosen canonical lives in a Search Console property you do not own, you cannot see the duplicate’s traffic. ⁷

Our own case: what our site asks for

The new seogenome.com emits a canonical tag on every page. ² I measured on September 14, 2026. Each page carries a single tag, inside the <head>, with an absolute URL. ¹ The address with a campaign parameter returns the canonical without the parameter. ¹ And the English page points to itself, with its Portuguese pair declared through hreflang. ^{1 2}

There was a stretch when that tag worked alone. On September 3, the www host served every inner page with a 200 response. The canonical pointed to the right address and was the only defense. ² The redirect arrived the same day, and the full story is in the piece on 301 redirects.

Now, the search from the opening. The two partners’ pages on the company site ask for something else. Their canonical points to each partner’s personal site. ¹ The rule has been in our site plan since August 31: the company’s page does not compete for the person’s name. ²

By the documentation’s yardstick, that request should not be granted. I compared the text of Daniel’s page on seogenome.com with the homepage of his personal site. Not a single sequence of eight words repeats between the two. ¹ And a canonical that does not resemble the page is never chosen. ¹¹

The search result fits a request that was turned down. It would also fit a request Google has not processed yet. Only URL Inspection tells the two apart, and I have not recorded that reading yet. ¹⁰

I found two more mismatches of our own that same day, and I fixed both. ¹ The sitemap listed the four partner pages, in Portuguese and in English. It suggested them as canonicals of themselves, while the tag pointed to another domain. That was the request down two paths that Google says to avoid. ⁶ Since September 14, 2026, pages with an external canonical stay out of the sitemap.

The other mismatch was on Polly’s page. The canonical pointed to pollymelo.com.br, which responds with a 301 to the www version. ¹ The RFC recommends against naming as canonical a URL that redirects permanently. ⁵ The tag now points to the final address, with www.

I leave the case on the table for you to weigh on your own site. This rule for the partners’ pages has an entity reason, and our site plan also expresses it in each page’s structured data. ² The canonical is the part of that rule that depends on the editor’s goodwill.

The afternoon check

Pick five important pages and a sheet with three columns. What the code asks for, what Google chose, and what you decide.

1. Open the code of each page and search for rel="canonical". Count how many show up. Check that the URL is absolute and that the tag sits inside the <head>. ^{6 9}
2. Open the address it points to. It has to respond with a 200, with no redirect, no noindex, and the same primary content. ^{5 9 11}
3. Add any parameter to the address, like ?test=1. The canonical has to stay the same, without the parameter. ⁶
4. Look for the five URLs in the sitemap. The listed address has to match the one in the tag. ⁶
5. Run URL Inspection on each one. Write down the declared canonical and the selected one. ¹⁰
6. In the Page indexing report, open the reason “Duplicate, Google chose different canonical than user.” Set aside the URLs that matter to the business. ¹¹
7. If the site has a paginated listing, open page 2. Check that its canonical points to itself. ⁸

Each mismatch becomes a line in the third column. The other signals that live in the page’s code are in on-page SEO. The rest of the technical work is in the technical SEO section.

The question you start asking

From now on, when someone tells you the canonical is set up, you will ask what Google chose. The code shows only the request. The answer lives in a field in Search Console.

The first step fits in today. Open URL Inspection on your most important page and compare the two canonical fields. If they differ, you have found your first refused request.