Pick the tool by asking whether the duplicate URL still needs to work for visitors. If it does not, a permanent redirect is the method Google documents for getting rid of duplicate pages [3]. If it does, rel="canonical" is the usual choice. noindex is rarely the right tool for choosing between duplicates, because it removes the page from Search rather than consolidating it [1].

This article covers the scenarios where people mix these up, and what Google says about each signal.

What each tool does

Google defines canonicalization as choosing a representative URL from duplicate or very similar pages. Duplicates can come from protocol, region, device, site-function, or accidental variants [2].

ToolWhat it tells GoogleDuplicate stays reachable?Strength per Google
Permanent redirectThe old URL has moved; use the new oneNo, visitors are sent to the targetStrong, and listed first [1]
rel="canonical"This URL is a duplicate of that oneYesStrong, listed second [1]
Sitemap inclusionThis URL is one we'd like indexedNot applicableWeak, listed third [1]
noindexKeep this page out of SearchYesNot a consolidation signal; blocks the page from Search [1]

Google calls redirects and rel="canonical" both strong signals. Sitemap inclusion is the weak one [1]. A signal is not a command, and Google can still choose a different canonical than the one you declared. The canonical tags guide covers how to check which URL Google actually picked.

Decision table by scenario

ScenarioUsual fixWhyWatch out for
http and https, or www and non-wwwPermanent redirect to one versionRedirects are the documented method for removing duplicate pages [3]Redirect directly to the final URL, not through a chain
Filter, sort or session parameters users needrel="canonical" to the clean URLThe variant must keep working, and Google prefers canonical over noindex for choosing a canonical within a site [1]Canonical is a signal, so Google may pick another URL
Parameters that serve no purpose (for example, old tracking links)Redirect, if the parameter's function is not neededRemoves the duplicate entirely [3]Redirecting can strip the parameter's function, such as analytics
Paginated listsEach page canonical to itselfGoogle says not to use the first page as the canonical [4]Do not point pages 2+ at page 1
A/B test variantsrel="canonical" to the originalGoogle recommends it over noindex for test pages [5]Keep the canonical on every variant
Syndicated copies on other sitesCase by case; see belowGoogle does not recommend canonical for syndication partners [6]You may not control the partner's markup
Staging or dev environmentAccess control first; see belowDifferent failure modes for each methodDo not rely on one method alone
Duplicate that must exist but stay out of SearchnoindexThe page is intentionally excludedIt will not consolidate signals into another URL

Canonical vs noindex

These are not interchangeable. Google says not to use noindex to prevent selection of a canonical page within a single site, because it completely blocks the page from Search. It prefers rel="canonical" for that job [1].

A noindex page is removed. A canonicalized page is folded into another URL as a duplicate. Choose noindex only when you want the page gone from Search, not merged into another.

The A/B testing guidance makes the same point. Canonical is recommended over noindex for test pages because it better matches the intent of keeping the original URL indexed [5].

Canonical vs redirect

Both are strong signals [1]. The difference is what happens to the visitor.

  • Redirect: the visitor and the crawler both end up at the target. Use a permanent redirect when you want to get rid of an existing duplicate. Google says all permanent redirection methods have the same effect on Google Search [3].
  • Canonical: the visitor stays on the duplicate. Use it when the duplicate has a job, such as a sort order, a printable view or a campaign landing variant.

Before redirecting a duplicate, check whether it has inbound links, bookmarks or external references. A redirect sends those visitors to the target, which is usually fine if the target is the same content. If it is not the same content, a redirect may be the wrong tool and you should look at the cases below.

Redirect behavior also matters to non-Google crawlers. The HTTP status codes guide covers what 301 and 302 communicate.

Pagination

Google says not to use the first page of a paginated sequence as the canonical page [4]. Pages 2, 3 and beyond list different items, so they are not duplicates of page 1. Pointing them all at page 1 tells Google to ignore the items they list.

The usual pattern is a canonical on each page that points to itself. That is general practice, not a quote from Google's documentation.

Syndication

Google's troubleshooting guide says the canonical link element is not recommended for avoiding duplication by syndication partners, because the pages are often very different [6]. That is Google's reasoning, and it applies to the syndication case specifically. A cross-domain canonical may be ignored if the syndicated page differs materially from the original.

Our notes from Google do not list a replacement method for syndication. If you syndicate, treat it as a decision to make with the partner. Settle whether their copy needs to be indexable at all, and what they are allowed to change. Do not assume a canonical tag on their page will do the work.

Staging and dev environments

Three different controls are in play, and they do different things:

  • Authentication (such as HTTP basic auth) means a request without credentials does not get the page content. It is a different category from the other two, which are requests to well-behaved crawlers rather than enforcement. See robots.txt is not a privacy control for why that distinction matters.
  • **noindex** asks Search not to index the page. It is a page-level directive, and it can only take effect if the crawler is able to fetch the page and read it.
  • robots.txt is a crawl instruction, not an indexing control and not a lock. If you use noindex, check that the same URLs are not also disallowed in robots.txt. A disallowed URL may not be fetched, and then the tag may not be seen.

A redirect from staging to production depends on how you use staging. If the hostname is an accidental leftover, a redirect may be reasonable. If your team works on staging, it will send them away from the environment they need. Decide that first, then choose the control.

After you fix it

Re-evaluation takes time. Google says duplicate clusters may persist for up to two weeks after you fix canonicalization issues [7]. Do not change the markup again in that window unless you find an actual error.

Redirects and canonicals both need consistent supporting signals. Internal links and sitemap URLs should point to the version you chose. See the technical SEO checklist for how to verify those.

What to do next

  1. List your duplicate URL patterns: protocol, host, parameters, pagination, test variants, syndicated copies, staging.
  2. For each, decide whether the duplicate must stay reachable for users.
  3. If not, apply a permanent redirect. If so, add rel="canonical" to the preferred URL.
  4. Reserve noindex for pages you want out of Search, not for choosing between duplicates.
  5. Check what Google selected in Search Console after a couple of weeks, not the next day.