What Is a Canonical Tag in SEO?
The most misunderstood thing in this section, because it looks like an instruction and behaves like a suggestion. Almost every problem with it comes from somebody expecting it to enforce something, which is worth establishing before anything else.
It Is A Suggestion, Not An Instruction
A canonical nomination tells a search engine which version of a page you consider authoritative. It is a strong hint rather than a command. Your choice is not guaranteed to be honoured.
What it states. A preference.
Among several addresses showing the same or similar content, this is the one you regard as the real page. That is a declaration of intent rather than an enforceable rule.
Why it can be overridden. The evidence may disagree.
Where other signals point elsewhere, else where the nominated page is substantially different from the one nominating it, a different version may be chosen. Your nomination is one input among several.
What people expect instead. Obedience.
The tag reads like a setting, so it gets treated as one. When the outcome differs from the nomination, people assume something is broken rather than that a judgement went the other way.
Why that expectation causes the damage. It invites overuse.
Somebody who believes this enforces something will use it to solve problems it was never meant for, which is the whole of block five.
What makes a nomination persuasive. Consistency.
A nomination agreeing with everything else your site says about that page is far more likely to be followed. That is block six, being the practical skill here.
Why Duplicates Exist At All
The same content commonly sits at several addresses without anybody intending it. That is normal rather than a fault. Understanding why prevents treating the symptom as misconduct.
Parameters. The biggest source.
Filters, sorting and tracking each produce a distinct address for the same content. URL structure covers how quickly these multiply.
Sorting and display options. The same page, rearranged.
A listing ordered by price, by name or by date is three addresses containing identical items. Nobody created three pages and three exist.
Print and alternate versions. The legacy case.
Separate versions built for printing or for a different context. Less common than they were and still present on established sites.
Syndication. Content elsewhere.
An article published on your site and also on a partner's. Two genuine copies at different addresses, where somebody has to be the original.
Why this is not a penalty situation. It is expected.
Duplicate addresses are a normal consequence of how sites work. Nothing is punished for having them. The nomination exists precisely because they are unavoidable.
Which Version To Nominate
Nominate the version you want people to reach. That single test resolves most cases. Where it does not, three secondary questions do.
The primary test. Where should somebody land.
If a customer clicked through from a search result, which address would you want them to arrive at. That is your canonical version. Everything else follows from it.
Is it complete. The first check.
The nominated version should contain the full content rather than a filtered or partial view. Nominating a subset means directing people to less than you have.
Is it stable. The second.
Prefer an address that will still exist next year. Nominating something temporary means the nomination breaks when it goes.
Is it clean. The third.
An address without tracking, session identifiers or filter values attached. Those variants exist for a purpose and none of them is being the definitive version of the page.
What we are not publishing. How it is written.
Where the nomination lives and how it is expressed is a development matter with a published specification. The decision is the part that needs a person who understands the business.
Pages Nominating Themselves
The normal state is for a page to nominate itself as the authoritative version. That sounds circular and it is doing real work, because it states an intention rather than leaving one to be inferred.
What it declares. This address is the real one.
The page is asserting that it is the definitive version of itself, which matters when variants of it appear with parameters or tracking attached.
Why silence is worse. Something else decides.
With no nomination, a judgement is made about which of several similar addresses is authoritative. That judgement might select a variant with tracking on it, which nobody wanted.
Why it sounds pointless. The circularity.
Nominating yourself reads as saying nothing. It is closer to signing your own work: the statement is not the information, the attribution is.
Where it should be the default. Nearly everywhere.
Most pages on most sites should nominate themselves. Nominating something else is the exception requiring a reason, rather than the other way round.
What platforms usually do. This, automatically.
Most modern platforms handle self-nomination without being asked, which is the right behaviour. The problems in block five come from somebody overriding it.
Misuse
Four failures, listed by consequence. Every one of them comes from expecting the nomination to enforce something, which is why block one comes first.
A page nominating a different page. The disappearing act.
A page that nominates something else is asking not to be treated as the real version. It can vanish from results entirely. This is the commonest cause of a page disappearing for no visible reason.
Every page nominating the home page. The catastrophic version.
Occasionally configured by somebody who misunderstood the purpose. The whole site is asking to be treated as one page, so everything except the home page can disappear.
Nominations contradicting redirects. The confusing one.
An address that redirects while also nominating something different, else two pages nominating each other. Both leave a decision open that you could have made, which is block six.
Nominations pointing at addresses that no longer work. The stale one.
Nominating a page that has been removed or moved. The nomination cannot be honoured and the page has still declined to be authoritative.
What they share. Silence.
Not one produces a visible symptom on the page. Every one of them is found by looking deliberately. The first two can remove pages from results without anything appearing to break.
Contradicting Yourself Is The Real Failure
A nomination saying one thing while a redirect, a sitemap or an indexing instruction says another leaves the decision to the crawler. Consistency is the whole job. This is the second of three places the same failure appears.
What the contradiction does. Hands away your decision.
Given two conflicting statements from the same site, something has to choose which to act on. You had a preference and you have declined to state it clearly.
The versions that occur. Four.
A nomination disagreeing with where a redirect sends people. A page nominated as authoritative while being excluded from a sitemap. A page nominating itself while instructing that it not be listed. And two pages each nominating the other.
Why it is worse than being wrong. Unpredictability.
A clear instruction you disagree with produces a predictable outcome. Two conflicting ones produce an outcome nobody can anticipate, which is harder to diagnose months later.
Where else this appears. Two other pillars.
A sitemap listing a page blocked elsewhere, plus international declarations that are not reciprocal. The same failure in a different mechanism.
What consistency requires. Reading your instructions together.
Checking what every mechanism says about the same page rather than each in isolation. That is rarely anybody's responsibility, which is why it rarely happens.
Pagination
A listing split across several addresses. The guidance changed. Canonicalising every page to the first is a common and damaging mistake rather than a matter of preference.
What changed. A signal was withdrawn.
Markup introduced in 2011 indicated the sequence of a paginated series. In 2019 Google announced it had stopped using it for indexing. The original announcement now carries a note stating it is no longer an indexing signal.
What the current guidance says. Self-reference.
For category pagination, each paginated address should be indexable and carry a nomination of itself. Google explicitly emphasises not nominating the first page from the others.
Why nominating the first page is damaging. The rest disappear.
Every page beyond the first is asking not to be treated as a real page, so the items on them can stop being found. On a large listing that removes most of the catalogue from consideration.
Why people do it anyway. It looks like duplication.
Pages two onward look similar to page one, so nominating the first seems tidy. They contain entirely different items, which makes them different pages rather than duplicates.
What actually helps a paginated series. Links.
Pages deeper in a sequence need internal links from somewhere other than the previous page. A chain reachable only by clicking through repeatedly is a discoverability problem no nomination solves.
What we are not publishing. The markup.
The decision is self-nomination on each page. How that is expressed is a development matter. A developer implementing it has the specification.
Canonicals Break During Rebuilds
These contain addresses, so any restructure leaves them pointing at the old ones. The site keeps nominating pages that no longer exist. Nothing on the site looks any different.
Why they break. They are written down.
A nomination names a specific address. Change that address and the nomination still names the old one, because nothing updates it automatically.
What that produces. A site nominating its former self.
Every page pointing at where it used to live. In the worst version the new pages are all declining to be authoritative in favour of addresses that no longer resolve.
Why it goes unnoticed. The new site works.
Pages load, the design is right and nothing reports a fault. This is the same shape of failure as the staging file in the robots pillar: invisible on the live site and checkable quickly.
When to check. Immediately after launch.
Whether pages nominate their current addresses, on a sample of each template. Ten minutes, catching a failure that otherwise takes months to surface.
Where the migration discipline sits. Its own cluster.
Our site migrations material owns canonicals during a move, alongside redirects and sitemaps. The full series is on the technical SEO guide.
A page can vanish
without anything
looking wrong.
A page nominating something else is asking not to be treated as the real version. Nothing on the page changes, nothing reports a fault, then it simply stops appearing. It is the commonest cause of a page disappearing for no visible reason.
What we check:
We publish the decision rather than the markup. A developer has the specification.
Every guide.
One practice.
How search finds and stores a site, what your own files are telling it, addresses and duplication, status codes, mobile, hosting and what a real audit contains.
