Short version: A canonical URL is the preferred URL that should represent a page when duplicate or very similar versions exist.
Canonicalization helps search engines understand which version of a page should collect signals and appear as the main version. It is often relevant when the same content is available through tracking parameters, filters, print versions, HTTP and HTTPS variants, or multiple category paths. A canonical tag is a signal, not a force field; search engines can still choose a different URL if the signals conflict.
Good canonical practice is consistent. The page should point to the preferred version, internal links should usually use that version, sitemaps should list that version, and redirects should not contradict it. Robots.txt blocks and URL removals are not proper canonical fixes because they solve a different problem.
Related terms: duplicate content, sitemap, and crawl budget.