Canonical Tags: The Five Mistakes That Undo Them

A canonical tag is a hint, not an instruction, and five common implementation faults make it easy to ignore. How to find them on your own site and what Google is telling you when it picks a different URL.

The canonical tag is the most misunderstood element in technical SEO, and the misunderstanding starts with its name. People treat it as an instruction. It is a hint.

Google evaluates your canonical alongside other signals, and where those signals conflict, it decides. When it decides against you, your reporting shows a page marked as a duplicate with a different canonical chosen, and most people read that as an error on Google’s part.

It usually is not. It is a report that your signals disagreed with each other. Here are the five ways that happens.

The prior question, of whether a canonical is even the right tool for your situation rather than a redirect or a noindex, is covered in robots.txt, noindex or canonical.

Mistake one: the canonical does not match the URL being served

The most common fault, and the most invisible.

A page is served at one address and declares a canonical at a slightly different one. The difference is usually a variant: http against https, with against without the www prefix, a trailing slash present or absent, uppercase characters in the path, or a tracking parameter left in.

Each of those is a different URL as far as a crawler is concerned. A page that says its canonical is the version with a trailing slash while being served without one is telling you one thing and itself another.

Most of the time this resolves harmlessly, because a redirect exists and the signals line up eventually. Sometimes it does not, and you get a page whose canonical points at a URL that redirects to a third URL, which is a chain of hints nobody can follow confidently.

The fix is to generate the canonical from the actual served URL rather than from a stored field, and to pick one variant of each dimension and enforce it everywhere.

Mistake two: there are two canonical tags

When a page declares more than one canonical, the sensible response from a crawler is to trust neither, and that is broadly what happens.

This almost always comes from two things generating tags independently. An SEO plugin and a theme. Two plugins. A plugin and a hand-coded tag somebody added years ago and forgot.

Nothing warns you. The page validates, the tags are individually correct, and the signal is simply discarded.

Check the rendered source of a few pages and count the canonical tags. We find duplicates often enough that it is worth doing before any deeper diagnosis, and on WordPress sites specifically it is one of the first things to look at.

Mistake three: JavaScript changes it

The raw HTML carries one canonical and the rendered page carries another, or the raw HTML carries none and a script adds it.

This matters more than it used to. Google renders, so it may well see the final version eventually. Most other crawlers do not render at all, so they see whatever is in the initial response, which in this case is either wrong or absent.

So the practical standard has tightened. The canonical must be correct in the raw HTML, in the head, on the first response. Treat anything injected afterwards as not existing.

The wider version of this argument, about what belongs in the first response and why, is in our piece on rendering budgets.

Mistake four: pagination canonicalised to page one

A paginated series where every page declares page one as its canonical. This was once common advice and it is wrong.

What you are saying is that pages two onwards are duplicates of page one. They are not. They contain different items. The result is that everything linked only from those later pages becomes harder to discover, because the pages that link to it have been declared redundant.

This is how a site ends up with half its catalogue or archive effectively invisible while every individual page looks correctly configured.

The right handling is for each paginated page to be self-canonical. It is its own page with its own content. If you do not want pages two onwards in the index, that is a separate decision expressed separately, and even then they should remain crawlable so the links on them still work.

The faceted and filtered version of this problem is covered in faceted navigation and pagination.

Mistake five: the canonical target is not a good page

A canonical pointing at a URL that redirects, returns a 404, is noindexed, or is itself canonicalised to somewhere else.

Each of these creates a contradiction. You are nominating a page as the preferred version while telling the crawler that page should not be indexed, or does not exist, or prefers a different page itself.

Faced with a contradiction, the crawler falls back on its own judgement, which is exactly the situation the canonical existed to prevent.

This creeps in over time rather than appearing at launch. A canonical target is deleted, or moved, or noindexed during a clean-up, and nothing checks the pages that pointed at it. It is a common finding after a migration, which is why it belongs on the migration checklist.

What Google is telling you when it disagrees

If your reporting shows pages where a different canonical was chosen, read it as diagnostic information rather than as a malfunction.

The common reasons, roughly in order.

  • Internal links contradict the canonical. You nominated one URL and link to another everywhere. Links are a strong signal and they outweigh a tag.
  • The sitemap contradicts it. You listed the non-canonical version.
  • The pages are not actually duplicates. You canonicalised two genuinely different pages together and the crawler can see they differ.
  • The nominated page is weaker. Fewer links, less content, or it is harder to reach.
  • Redirects point the other way. You redirect towards the page you declared non-canonical.

In almost every case the fix is to make your other signals agree rather than to argue with the tag. Link to the canonical version internally, list only the canonical version in your sitemap, and make sure redirects point the same way.

The duplicate-content myths around all of this are addressed in handling duplicate content.

Cross-domain canonicals, briefly

You can canonicalise to a URL on another domain, which is the right tool for syndicated content where somebody else republishes your article and points back to you.

Two cautions. You are dependent on the other party implementing it and not removing it later. And it is a hint across domains just as it is within one, which means it is weaker still when other signals point elsewhere.

Do not use it to consolidate two of your own sites. That is a migration, and a redirect is the correct instrument.

How to audit this on your own site

  1. Crawl your site and export the canonical for every URL. Compare it against the URL itself. Every mismatch is worth a look.
  2. Count canonical tags per page. Anything above one is a bug.
  3. Fetch a sample without JavaScript and confirm the canonical is present and correct in the raw response.
  4. Check every canonical target’s status. Any that redirect, 404 or are noindexed need fixing.
  5. Check your paginated series. Each page should be self-canonical.
  6. Compare your sitemap against your canonicals. The sitemap should contain canonical URLs only.
  7. Spot-check internal links for the pages where Google disagreed with you.

Then add a check to your deployment so a theme or plugin update cannot silently introduce a second tag, which is how most of these problems arrive.

What is a waste of effort

Canonicalising everything to the home page to consolidate authority. It does not work that way, the pages are not duplicates, and the signal will be ignored.

Using a canonical where a redirect is correct. If a page has genuinely moved, redirect it. A canonical leaves the old URL live and reachable.

Using a canonical to hide a page you do not want indexed. That is what noindex is for, and the two express different things.

Arguing with a chosen canonical in reporting without checking your own signals first. The report is usually right about the disagreement even when you dislike the outcome.

Adding canonical tags to pages that have no duplicate situation and then forgetting them. A self-referencing canonical is good practice. One pointing somewhere you no longer remember is a trap for the next person.

The short version

Generate the canonical from the served URL. One tag per page. In the raw HTML. Pointing at a page that is live, indexable and self-canonical. Supported by internal links, sitemap entries and redirects that all agree.

Do that and the tag does its job quietly, which is the only way you will ever know it is working.

If your reporting is full of pages with a different canonical chosen and you cannot work out why, get in touch and we will trace the signals.