In short
Partly. A canonical tag tells Google which version of a page you consider the original, and when Google agrees it consolidates the ranking signals onto that URL. It is a hint rather than a directive, Google regularly selects a different canonical, and the tag does nothing to stop the duplicate addresses being crawled in the first place. Use canonicals to tidy the duplicates a store cannot avoid, and fix the URL generation itself where the duplicates are being created by your own site.
The tag is a recommendation, and Google decides
Google treats the canonical element as one input among several when it picks which URL from a set of similar pages to keep in its index. The others include internal linking, which version appears in sitemaps, redirects, hreflang relationships and which URL looks more useful. If those signals contradict the tag, the tag can lose.
That is why a store can have a correct canonical on every parameter address and still see the parameter version ranking. The usual reason is that the site itself links to the parameter version everywhere, so the strongest internal signal disagrees with the tag.
The practical consequence is that canonicals work best when everything else agrees with them. Internal links point at the canonical URL, the sitemap lists only canonical URLs, the canonical page returns a normal response, and the pages really are equivalent. A canonical pointing from a page to a substantially different page is likely to be ignored.
It also has no effect on crawling. Both addresses continue to be requested, which matters on a large catalogue where crawl attention spent on duplicates is attention not spent on new products.
What canonicals genuinely resolve on a store
They are the right tool for duplicates that are a normal part of ecommerce and cannot be engineered away. Tracking parameters appended by email and ad platforms. Sort and view parameters on a category page. Session or currency parameters that some setups add. Variant selection parameters that produce one address per option. Product pages reachable through a category path as well as at their own URL, which Shopify canonicalises for you by default.
They also help with near-duplicate products that genuinely should be one page, such as the same item listed under two part numbers by different suppliers, and with print or quick-view versions of a page where those exist.
What makes these cases suitable is that the content is the same and the alternative addresses arrive from outside your control. You are telling Google which of several equivalent doors is the front one.
One detail worth getting right: the canonical URL should be absolute, should use the protocol and hostname you actually serve, and should be self-referencing on the canonical page itself. Themes and apps sometimes emit a second canonical or overwrite the first, and two conflicting tags on a page cause Google to ignore both.
What they will not do for you
They will not control crawl budget. If faceted navigation is producing hundreds of thousands of combinations, canonicals let Google know which one counts but do not stop it requesting the rest. That problem is solved by not linking the combinations, by robots rules, and by deciding which facets are allowed to exist as URLs at all.
They will not merge pages that should have been merged. Two thin product pages for what is really one product remain two thin pages, and the better fix is a redirect and a single stronger page.
They are not a substitute for a redirect when a URL has permanently moved. A canonical keeps both addresses live and reachable; a 301 moves the shopper and the signal in one step, and it is the correct answer for retired products, changed handles and migrated structures.
They should not be used across paginated series. Pointing page two onward at page one tells Google the deeper pages are duplicates of the first, and the products only listed on those pages can lose their discovery path. Let each paginated page be canonical to itself.
Finally, they do not fix duplicate content on other people's domains. Syndicated descriptions on marketplaces and retailer sites are a distinctness problem for your product copy, not something a tag on your own page can resolve.
Checking that yours were accepted
Use the URL inspection tool in Search Console on a duplicate address and read the two lines it reports: the canonical you declared and the canonical Google selected. When they match, the tag is doing its job. When they differ, Google has weighed the other signals and reached its own conclusion, and the message tells you which URL it chose instead.
The page indexing report gives the same picture at scale. The category described as a duplicate where Google chose a different canonical is the one to work through, and the pattern in those URLs usually names the cause: a parameter you did not know was being generated, a collection path in internal links, or a locale version competing with the original.
From there the fix is almost always upstream. Change the links so they point at the canonical URL, remove the duplicate addresses from sitemaps, and stop the site emitting them where you can. The tag then only has to handle the addresses that arrive from outside.
Recheck after a few weeks. Canonical selection is not instant, and on a large catalogue the reports move gradually as pages are recrawled.
Related questions
Why does Search Console say Google chose a different canonical?
Because the other signals on your site pointed somewhere else. The usual causes are internal links to the version you did not choose, the wrong URL in the sitemap, a redirect chain, or two pages that are similar enough for Google to prefer the one it considers more useful. Align those signals and the selection usually follows.
Should page two of a category canonicalise to page one?
No. Each paginated page should carry a canonical pointing to itself, so the products listed only on later pages keep a route to discovery. Pointing them all at the first page suggests the rest are duplicates and can leave part of the catalogue reachable only through search.
Do canonical tags pass link value from the duplicate to the original?
When Google accepts the tag, it consolidates the signals it associates with those URLs onto the chosen one, which includes links pointing at the duplicate. That consolidation is not guaranteed, and it is another reason to prefer a redirect where a URL has genuinely moved rather than leaving two addresses live indefinitely.