Canonicals, redirects and duplication
When several addresses carry the same content, the engine keeps one and groups the others. You can indicate which one by several means, including the permanent redirect and the canonical tag. These are signals, not instructions: the final decision is not yours.
Why does an engine group addresses?
Because the same content is nearly always reachable by several paths, and there is no point storing it more than once. A site produces duplicates without meaning to: with and without a trailing slash, over HTTP and HTTPS, with a tracking parameter, with a sort order, in a printable version. Each of those variants is a distinct address to a crawler.
So the engine picks one, treats it as the representative of the group, and attaches the others. That mechanism is not a punishment: it is tidying. The problem starts when the representative kept is not the one you would have chosen.
How do you declare the address you prefer?
To name the canonical address of a duplicate or very similar page, several methods exist and combine: the canonical link tag, an HTTP header, the sitemap and the permanent redirect.
officialSource : Google Search Central, “How to specify a canonical URL”
The important word is “methods”, in the plural. There is no switch but a bundle of signals, and they all have to tell the same story. A canonical tag naming one address, a sitemap declaring another and internal links pointing at a third produce a site that does not know what it wants — and a decision taken without it.
Permanent or temporary redirect: the only difference that counts
Both take the visitor to the same place. To an engine they nonetheless do not carry the same weight, and the documentation is precise on the point.
The crawler follows a permanent redirect, and the indexing pipeline uses it as a signal that the target should be canonical.
officialSource : Google Search Central, “Redirects and Google Search”
The crawler follows a temporary redirect too, but the indexing pipeline does not use it as a signal that the target should be canonical.
officialSource : Google Search Central, “Redirects and Google Search”
Two almost identical sentences, one negation apart, and all the practical difference in the world: the permanent one passes the canonical signal on, the temporary one does not. A rebuild carried out with temporary redirects therefore leaves the old address a candidate for the index, indefinitely.
The advice that comes with that distinction is wiser than it looks:
The documentation advises using a permanent redirect when you are certain it will not be undone.
officialSource : Google Search Central, “Redirects and Google Search”
It moves the question from the technical to the organisational: the right redirect is not the one that is correct in protocol terms, it is the one matching a decision you do not expect to reverse. A three-week commercial operation is not a change of address.
Aligning every canonical signal on a single address
hypothesis
- Effort
- moderate
- Time to effect
- weeks
- Risk
- low
- Durability
- structural
On this site, measured on 2026-08-16, the address in http with www reaches the version served through two successive redirects, where the three other forms need one or none: the move to TLS and the removal of the subdomain are handled by two distinct rules.
testedProtocol : Request each form of the address without automatically following redirects, and note every Location header until the final response. Ten addresses measured on 2026-08-16; full table in section C.2 of the public register.
What do you do when the engine does not follow your declaration?
It happens, it is foreseen, and it has a name:
The indexing report publishes a state for duplicate pages where the engine kept a canonical address different from the one the site declared.
officialSource : Google Search Console, “Page indexing report”
Observing that state is useless if you read it as a malfunction. It reads as information: the site’s signals did not convince. The most frequent causes are prosaic. The page declared as canonical redirects elsewhere. Internal links point overwhelmingly at the variant. The sitemap declares the attached address and not the representative. Or the two pieces of content are in fact so close that the choice between them is arbitrary, in which case the real question is why two pages exist.
Update of 11 August 2026
The canonicalisation troubleshooting guide was clarified on 10 July 2026: the engine now states that pages can stay grouped for up to two weeks after content problems are fixed. That duration was missing, and its absence produced a known mistake — undoing a correct fix, too early. The full record is in the monitoring log.
The case of parameters and versions
A tracking parameter added to an address manufactures a new one. The most economical rule is therefore not to produce any in internal links: an internal link never needs a campaign parameter. For functional parameters — sort, filter, pagination — the decision is taken page by page, and it joins up with crawl budget: you have to know which of those variants deserve to exist as pages, and which are only display states.
Finally, a useful reminder after any migration:
After a redirect, the old addresses can still appear as alternative names in the results, and this fades without intervention.
officialSource : Google Search Central, “Redirects and Google Search”
So there is nothing to do during that period, and it is precisely the moment when the temptation to act is strongest.
What does not work, or no longer works
Put a self-referencing canonical tag on all your pages, and a canonical to the main page on your similar ones: Google will always respect your choice.
One last reflex is expensive: chaining redirects across successive rebuilds, until you have runs of three or four hops. The indexing report treats a redirect error as a failure cause in its own right, and a long chain eventually becomes one. After every migration, the only check that counts is that an old address reaches its final target in a single hop.
Sources
- How to specify a canonical URL with rel="canonical" and other methods — Google Search Centralaccessed on
- Redirects and Google Search — Google Search Centralaccessed on
- Page indexing report — Google Search Consoleaccessed on
- Learn about sitemaps — Google Search Centralaccessed on
Block revalidated on
Frequently asked questions
- Is a canonical tag an instruction?
- No, it is one signal among several. The indexing report even publishes a state for pages where the engine kept an address different from the one declared. When that happens, it means the site's other signals contradicted the declaration.
- What is the difference between a permanent and a temporary redirect?
- Both are followed. Only the permanent one acts as a signal that the target should become the canonical address. The temporary one moves the visitor without passing that signal on, which suits a short operation and does not suit a definitive change of address.
- How long should a redirect be kept?
- The documentation sets no duration, it advises using the permanent form only if you are certain you will not go back. In practice a redirect is kept as long as external links point at the old address, and that means a long time.
- Why does my old address still appear in the results?
- This is documented and temporary: after a redirect, old addresses can survive for a while as alternative names, then fade on their own with no intervention from you. There is nothing to fix and above all nothing to force, since haste only multiplies contradictory signals.