The citation still works. The evidence has changed.

An investigator returns to a paper cited by an article and finds that its landing page has moved. The DOI resolves, so the citation appears healthy. But the paper has been corrected, the supplementary files are elsewhere, or the original dataset is no longer available. A working link was never a complete record of the evidence.

Persistent identifiers are an enormous improvement over casual URLs. They are not magic containers that freeze the intellectual state of everything they identify.

The DOI Handbook^{[1]DOI Handbook} describes persistent identification through metadata, resolution and governing organizations. A DOI can continue to point to information even when a resource moves. That continuity depends on registration agencies, registrants and ongoing upkeep; it is not produced by a checksum in the identifier itself.

A citation can name a work while failing to capture the precise version used by an author. Corrections, retractions, linked datasets and supplemental material may change how a reader should interpret the original claim.

For a calculation, the strongest record may include the version of a paper, exact table or page, dataset version, transformation code and date accessed. If the argument depends on a reported numerical estimate, a general DOI to the latest article is not enough to reproduce the value.

There is also a distinction between identifying an object and guaranteeing public access to it. An identifier can be stable even when licensing or repository policies limit distribution. An archive may retain a lawful copy for preservation without giving every visitor permission to republish it.

Suppose an article argues from the result in Table 2 of a study. The publisher later issues a correction to that table. The responsible publication should not silently replace its historical wording as though the original author had always read the corrected version.

Instead, retain the original citation and observation date, publish a clearly dated correction or update, and link to the changed material. The source relationship becomes richer: observed version, superseding version, editorial consequence.

An independent newsroom needs more than outbound links. It needs source capture policies, rights-aware preservation, checksums where appropriate, correction history, and a way to distinguish an original observation from later copies. The Aurekai essay on dependent evidence shows why multiplying citations does not automatically multiply independent support.

The design goal is not to make every paragraph an academic paper. It is to let a reader follow the chain from a claim to the versioned material that justified it, then see when and why that justification changed.