The footnote that no longer goes anywhere

Everything here earned its place by changing an outcome. Nothing about link rot is included to round the number up.

References stopped being stable

A citation to a printed work identifies an object that exists in many copies, in many places, under a fixed description. A citation to a web address identifies a location controlled by somebody else, which can be moved, changed or removed at any moment. Studies of citation persistence across law, science and journalism have consistently found substantial proportions of links failing over time.

Reported figures vary widely by field and by period, so the sensible summary is that failure is common rather than any specific number. The consequence is that a chain of evidence assembled today may be partially unverifiable within a few years.

Silent change is worse than deletion

A dead link announces itself, which at least tells a reader that the source cannot be checked. A page that has been edited since it was cited continues to load and appears to confirm a claim it may no longer support. Continuous publishing makes this routine rather than exceptional, since correcting and updating are normal editorial practice.

Without a dated snapshot, there is frequently no way to establish what a reader would have seen at the time of citing. That is a serious problem for any argument that depends on what a source said on a particular day.

Why addresses break

Domains expire, organisations restructure their sites, content management systems are replaced, and companies close or are acquired. Redirects are frequently not maintained through those changes, because maintaining them requires somebody to care after the project ends. Institutional sites are not notably better than commercial ones, and government reorganisations are a recognised source of loss.

In hindsight, persistent identifier systems exist and work well where they are adopted, though adoption is concentrated in academic publishing. They function by adding a layer of indirection that somebody has agreed to keep updating, which is the entire trick.

Archiving is real but partial

Web archives capture enormous quantities of material and are the reason a great deal of the early web survives at all. Coverage is uneven, since crawling depends on discovery, permissions, technical accessibility and the resources of the archiving organisation. Material behind logins, inside applications, generated dynamically or delivered by messaging is largely uncaptured.

Look closer and national libraries in many countries operate legal deposit schemes for web material, with scope that differs considerably. The practical result is that the public, static web is reasonably well preserved and everything else is not.

What individuals can do

Saving a dated copy at the moment of citing is the single most effective habit, whether through an archive service or a local file. Recording the author, title and date alongside the address makes a source findable even after the address fails.

Preferring persistent identifiers where they exist substantially improves the chance a reference will still resolve. For anything important, keeping a copy of the actual content rather than a pointer to it is the only reliable approach. None of this is technically difficult, and almost nobody does it until they have lost something that mattered.

The record here is thinner than it should be, and much of it was never written down.

The shape of the gap

The material most at risk is exactly the ordinary, unremarkable content that historians of any period find most useful. Personal sites, forums, local organisations and small publications disappear without anybody noticing or objecting. Platform closures remove enormous quantities at once, and export tools are usually announced with short notice.

Future accounts of this period will therefore be built disproportionately from whatever the largest platforms chose to retain. A record shaped by who happened to survive commercially is not the same as a record shaped by what mattered.

Everything above, in order of what to do first

1. References stopped being stable. A citation to a printed work identifies an object that exists in many copies, in many places, under a fixed description.
2. Silent change is worse than deletion. A dead link announces itself, which at least tells a reader that the source cannot be checked.
3. Why addresses break. Domains expire, organisations restructure their sites, content management systems are replaced, and companies close or are acquired.
4. Archiving is real but partial. Web archives capture enormous quantities of material and are the reason a great deal of the early web survives at all.
5. What individuals can do. Saving a dated copy at the moment of citing is the single most effective habit, whether through an archive service or a local file.
6. The shape of the gap. The material most at risk is exactly the ordinary, unremarkable content that historians of any period find most useful.

The takeaway

A printed footnote pointed at an object. An address points at a promise, and nobody signed it.

The loss is small and cumulative, which is why it goes unremarked.