What a canonical tag actually does
A canonical tag tells search engines which URL you consider the real address of a piece of content, when several addresses show the same thing.
<link rel="canonical" href="https://example.com/blue-widgets" />That line, in the <head> of a page, says: *this content's proper home is that URL; consolidate the signals there.*
The critical word is consider. A canonical is a hint, not a directive. Google treats it as one signal among several — alongside internal links, sitemaps, redirects and which version is linked from elsewhere — and decides for itself. If everything else on your site contradicts the canonical, Google can and will ignore it.
That single fact explains most canonical problems people report. The tag is not broken. It is being outvoted.
When you need one
Duplicate URLs are usually created by accident, not by writing the same article twice. The usual sources:
- Tracking parameters.
/productand/product?utm_source=newsletterare different URLs with identical content. - Faceted navigation. Filters and sort orders multiply one category page into dozens.
- Session or pagination artefacts.
?page=1alongside the bare URL. - Protocol and host variants.
http://,https://,www.and non-www.versions. - Syndication. The same article republished on another domain.
For most of these the canonical points at the clean version and the problem is contained.
It is also worth adding a self-referencing canonical to every indexable page — a canonical that points at itself. It costs nothing, and it forecloses a whole category of accidental duplication from parameters you did not anticipate.
Canonical, redirect or noindex — which one
These three get confused constantly, and the wrong choice is expensive. They do different jobs.
| Tool | Use it when | What happens |
|---|---|---|
| 301 redirect | The duplicate should not exist at all | Visitors and crawlers are sent to the real URL; the old one disappears |
| rel=canonical | Both URLs must remain reachable | Both stay live; ranking signals consolidate on the canonical |
| noindex | The page is useful to people but not to search | Stays reachable, removed from the index |
The decision is usually straightforward: if a human still needs to reach the duplicate URL — a filtered view, a tracked link, a printer-friendly version — canonicalise it. If nobody needs it, redirect it.
Never combine noindex with a canonical pointing elsewhere. You are saying "this page is a copy of that one, consolidate onto it" and "remove this page" simultaneously. The instructions conflict, and the outcome is unpredictable — sometimes the target inherits the noindex.
The five ways canonicals go wrong
1 · Every page canonicalises to the homepage. A plugin misconfiguration, usually. The effect is that the entire site claims to be one page, and the rest can drop out of the index. If a large number of pages have vanished at once, check this first — it takes ten seconds and it is the most destructive of the five.
2 · The canonical points at a page that is blocked, redirected or 404. The target has to be reachable and indexable. A canonical pointing at a URL blocked in robots.txt cannot be verified, so it is ignored — and you are left with duplicates and no consolidation.
3 · Relative URLs, or the wrong protocol. href="/blue-widgets" works on some setups and resolves unexpectedly on others. Always absolute, always including https:// and the exact host you actually use. A canonical pointing at the http:// version of your own site is a common and quiet error.
4 · Multiple canonicals on one page. A theme adds one, a plugin adds another. Two conflicting tags mean neither is trusted, and Google falls back to its own judgement. This shows up most often where an SEO plugin and a page builder both think they own the <head>.
5 · Canonicals that contradict everything else. Your canonical says page A, your internal links all point to page B, your sitemap lists B, and B is what other sites link to. Google follows the weight of evidence, not the tag. A canonical only works when the rest of the site agrees with it.
Paginated pages and faceted navigation
Two specific cases worth stating, because the received advice has changed.
Pagination. Do not canonicalise page 2, 3 and 4 of a series back to page 1. Those pages contain different items, and consolidating them tells Google the deeper items do not exist as distinct content. Each page in a sequence should carry a self-referencing canonical. If you also offer a genuine view-all page, canonicalising the sequence to that is defensible — but only if it really does contain everything.
Faceted navigation. Filter combinations can generate enormous numbers of URLs. Canonicalising every combination back to the unfiltered category is the usual reflex and is often right — but only for combinations with no independent search demand. If people genuinely search for "waterproof walking boots size 9", that filtered page may deserve to be indexed in its own right. The decision is per facet, not site-wide.
Cross-domain canonicals
A canonical can point at a URL on a different domain. This is how syndication is supposed to work: the republishing site points its canonical at your original, and the ranking signals stay with you.
Two cautions. It relies entirely on the other site implementing it, which you cannot control or verify without checking. And because a canonical is a hint, a large, authoritative republisher can still outrank the original — Google weighs its own signals.
If retaining the rankings matters, a cross-domain canonical is a request, not a guarantee.
How to check what is actually set
- View the page source and search for
rel="canonical". Check there is exactly one, that it is absolute, and that it points where you expect. Do this on the rendered page too if the tag is injected by JavaScript. - URL Inspection in Google Search Console shows both the canonical you declared and the one Google selected. When those differ, Google has overruled you, and the report is the fastest way to find out.
- Crawl the site with any crawler that reports canonicals, and sort by target. A canonical pointing somewhere unexpected stands out immediately at scale in a way it never does page by page.
- Check the HTTP header version. Canonicals can also be set as an HTTP
Linkheader, which does not appear in the HTML source. PDFs and other non-HTML files use this method, and a header canonical conflicting with an HTML one is genuinely hard to spot.
Frequently asked questions
Is a canonical tag a directive or a hint?
A hint. Google treats it as one signal among several and may select a different canonical — internal links, sitemaps and external links all count. Search Console's URL Inspection shows which URL Google actually chose.
Should every page have a canonical tag?
A self-referencing canonical on every indexable page is good practice. It costs nothing and prevents accidental duplication from tracking parameters you did not anticipate.
Can I use a canonical instead of a 301 redirect?
Only when both URLs need to stay reachable. If nobody needs the duplicate, redirect it — a redirect is a directive and a canonical is a hint, so the redirect is the stronger and cleaner signal.
Why is Google ignoring my canonical tag?
Usually because the rest of the site contradicts it: internal links, the sitemap or external links all point at a different URL. Other causes are a canonical target that is blocked, redirected or returns a 404, and more than one canonical tag on the same page.
Get your canonicals set correctly
Canonical problems are quiet. Nothing errors, nothing looks broken, and pages simply stop appearing — often months before anyone connects the two.
If you would rather have this verified than assumed, the Technical SEO Starter Setup covers canonicals, crawl directives and indexation together — $497, one-off, no retainer.