Canonical tag

A canonical tag names the address that should be treated as the real version of a page. It is normally one line in the head of the page, and for files that are not web pages it can be sent as a header instead. It exists because the same content often sits at more than one address, usually without anybody intending it, and without the tag a search engine has to guess which one you meant.

In short

• It names the preferred address when several show the same content.
• Duplication is not penalised. Google just picks one and ignores the rest.
• The tag is a strong hint, and Google can decide otherwise.
• Pointing everything at the homepage is either useless or destructive.

Where the duplicates come from

Almost nobody creates duplicate pages deliberately. They accumulate out of how websites and links work.

• Parameters. A filter, a sort order or a tracking tag appended to a link produces a new address showing the same page.
• www and the bare domain. Both usually load. To a machine reading them literally, they are two sites.
• http and https. The same again, if the old version was never redirected.
• Trailing slashes. An address ending in a slash and one without it are technically different.
• Products in several categories. A shop that files one item under two departments often gives it two addresses.
• Pagination and print views. Older platforms in particular still generate these quietly.

A small shop can produce hundreds of these without a single page being written twice.

What Google does about it

Left alone, Google groups the addresses it thinks are the same, chooses one, and indexes that. The others stay reachable for visitors and stop appearing in results. This is worth restating because the received wisdom is more frightening than the reality. There is no duplicate content penalty for ordinary cases like the ones above. Nothing is deducted and nobody is punished. The cost is only that the choice was made without you, and the address Google settles on may be the one with a tracking tag on the end, or the print version, or the one nobody links to.

The tag is how you get a say.

It is a hint, not an order

This catches people out more than anything else here, and it is the reason a correct-looking tag sometimes changes nothing.

Google treats the tag as one signal among several. Your internal links are another. Your sitemap is a third. Redirects are a fourth, and the strongest. When those agree, the tag is almost always honoured. When they disagree, for instance when every link on your site points at one address while the tag names another, Google resolves it and may well side with the links.

So the useful way to think about a canonical tag is as a vote you cast rather than a switch you flip. If you want the vote to carry, make the rest of the site agree with it.

The mistakes that cost pages

Four of these turn up often enough to be worth naming. They are listed in rough order of how much damage they do.

• Everything pointing at the homepage. Usually a template edit copied to every page. Most of the time Google can see the pages are not duplicates of the homepage and ignores the instruction, so the tag achieves nothing. When it is taken at face value, pages drop out. If organic traffic falls off a cliff after a redesign, check this first.
• Canonical plus noindex on one page. Google advises against the combination outright. A noindex removes the page from search altogether, which is a far blunter outcome than the consolidation the canonical tag was asking for. Choose one, and for duplicates it should be the canonical.
• Pointing at a page that redirects or is missing. The signal has nowhere to land, so it gets ignored.
• Tags that disagree with the sitemap. Only addresses you have declared as canonical belong in an XML sitemap. Listing the others sends two messages.

Not the same problem as two pages competing

There is a distinction here that is easy to lose, and losing it leads to the wrong fix being applied. Canonical tags are for one page living at several addresses. The content is identical or near enough, and nobody meant to publish it twice. The tag says these are the same thing, keep this one.

Two genuinely different pages both written to win the same search is a separate problem, described under keyword cannibalisation. A canonical tag does not fix that, because the pages are not duplicates. They are two real pages that need one of them repurposed or merged. Reaching for a canonical tag there usually removes a page that was earning something.

Words you will hear

• rel=canonical. The attribute itself, written into the head of the page.
• Self-referencing canonical. A page naming its own address. The sensible default.
• Canonicalisation. The whole process of deciding which address represents a page.
• Google-selected canonical. The address Google chose, which Search Console will show you next to yours.
• Parameter. Anything after a question mark in an address. The most common source of accidental duplicates.
• noindex. An instruction to keep a page out of results. Not a substitute for a canonical tag, and not to be used alongside one on the same page.

What to check today

Open your own site and load the homepage four ways: with www and without, with https and http, and with something harmless appended after a question mark. If all four load and none of them redirects, look at the source of each and see whether the canonical tag names the same address every time.

Then open the page indexing report in Search Console and look for pages where Google chose a different canonical than you did. That list is short on most sites and it is the only place the machine tells you plainly that it disagreed with you.