A citation can expire while the claim remains

A well-sourced essay can become poorly sourced without changing a word. The agency reorganizes its website. A standards body publishes a new edition. A company removes an old announcement. A deep link begins returning an error, redirecting to a generic homepage, or serving a different document at the same path. The claim remains in the essay, but the route by which readers could examine its support has decayed.

This is usually treated as a maintenance nuisance. It is more serious for arguments about law, policy, science, and technology. A citation is part of the reader’s ability to test the writer’s account. When it fails, the text may still be correct, yet the reader must accept more on trust. Link decay quietly changes the epistemic terms of publication.

The web was designed to tolerate change, not to guarantee permanent location. HTTP can report that a resource moved, disappeared, or is temporarily unavailable. Redirects can preserve continuity, but publishers control them unevenly and organizations often rebuild sites without maintaining old paths. A URL is an address under another institution’s custody.

A successful response is not enough

Link checking often stops at the status code. A response of 200 means a representation was returned; it does not prove that the representation still supports the claim. A dead report may redirect to a department homepage. A removed product page may become a marketing landing page. A path may be reused for a revised document whose conclusions differ from the cited edition.

Maintenance therefore needs several levels. Transport checks detect unreachable links and redirect loops. Content checks verify titles, issuing organizations, dates, document identifiers, or expected phrases. Editorial review asks whether the source still supports the sentence. The first can be automated broadly. The last remains an act of judgment.

A citation is healthy only when a reader can still reach the evidence that the sentence asks it to carry.

Source selection can reduce decay. Prefer the issuing institution’s stable document page over a news summary, an official standard over a vendor paraphrase, and a persistent identifier over a transient search result. The HTTP specification itself is available through a durable RFC identifier as well as working-group pages. Durable naming does not eliminate change, but it makes custody clearer.

References should preserve enough metadata to recover from movement: source title, issuing body, publication date, document number, and perhaps the relevant section. A bare phrase linked to “here” gives a maintainer little to search when the URL fails. A descriptive citation can be repaired even after the original route disappears.

Repair the route without rewriting history

When a source moves, updating the link to the same official document is ordinary maintenance. When the source has been superseded, the choice is more delicate. The essay may have made a historically accurate claim based on the version available at publication, while current readers also need to know that the rule changed. Silently replacing the citation with a newer source can make the old sentence appear to claim something it never did.

A useful repair distinguishes historical support from current status. Keep the original edition identifiable, link to an official archive if available, and add a dated note when the change materially affects the argument. The goal is not to keep every sentence looking eternally current. It is to preserve what was knowable then while helping readers understand what is true now.

Web archives can help, but they are not a universal answer. Some material cannot be archived lawfully or completely. Interactive pages may not survive. An archived copy may lack the issuing institution’s current authority. Use archives as evidence of past publication, not as a substitute for evaluating provenance.

The corpus should make link health observable. Scheduled checks can produce a queue of failures with the referring essay, anchor text, prior destination, current response, and first observed date. Group failures by domain to distinguish a temporary outage from a site-wide migration. Prioritize citations carrying central factual claims over optional reading.

Repairs deserve review because links are editorial content. A replacement source may be weaker, partisan, or only superficially similar. Automated search can propose candidates; a reviewer should confirm the issuing body, version, and support. The record should note the change so later maintainers know why the URL differs from the original release.

Persistent identifiers are worth seeking where a domain supports them. Digital object identifiers, RFC numbers, court citations, standards identifiers, and agency publication numbers decouple recognition of a document from one website’s current folder structure. The W3C’s note on cool URIs captures the broader principle that stable names require institutional commitment. A writer cannot manufacture that commitment, but can prefer sources that have made it.

Writers can also design claims for resilience. Quote sparingly, paraphrase accurately, name the source in the prose, and avoid making a fragile external page carry more certainty than it contains. Where a claim depends on a specific version, say so. Where the source is illustrative rather than dispositive, make that role clear.

A publication accumulates obligations as it accumulates links. The obligation is not to freeze the web. It is to keep the path from claim to evidence usable and honest. Citation maintenance is therefore part of publishing, not janitorial work after publishing. Budget time for it alongside new writing, and publish material repairs transparently. Review the evidence, not merely the address. The argument remains answerable only while readers can inspect what it rests on.

— Dispatches · Summit Cognitive