The Internet Never Forgets—Until It Does

There is a warning almost everyone who grew up with the internet has heard:

Once it’s online, it’s there forever.

Usually, it is good advice.

A photograph can resurface years later.

An old comment can be found.

A screenshot can survive long after the original disappears.

The internet can be remarkably good at preserving copies of things we wish it had forgotten.

But try returning to the internet you actually used ten or fifteen years ago.

Old bookmarks lead nowhere.

Images are missing.

Videos no longer play.

Forums have vanished.

Accounts have disappeared.

Links inside articles point to pages that changed—or stopped responding entirely.

The internet can preserve a single embarrassing photograph for decades while losing the website, conversation, and community that once made it understandable.

That is because digital permanence and digital preservation are not the same thing.

When we imagine preserving something online, we usually think about the content itself.

A photograph.

A post.

A video.

A document.

But on the web, content is only one layer.

Access also depends on an address that still leads somewhere.

Researchers studying citations in D-Lib Magazine examined 4,387 unique URLs referenced in 453 articles published between 1995 and 2004.

By their final observation in 2005, about 30% of those URLs no longer resolved. Based on the pattern they observed, the researchers estimated a roughly ten-year half-life for URLs in that collection.

The articles could survive.

The references around them could decay.

A link is not merely directions to information.

It is part of the structure connecting one piece of information to another.

When the link disappears, the page may remain readable while part of its evidence vanishes.

There is another kind of decay that is harder to notice.

The URL still works.

The page still loads.

Nothing appears broken.

But the information at that address is no longer the information the original author meant to cite.

Researchers examining links in Harvard law journals and United States Supreme Court opinions distinguished this from ordinary link rot.

They called it reference rot.

In their sample, more than 70% of URLs in the examined law journals and about half of URLs in Supreme Court opinions suffered from either link rot or reference rot.

This creates a strange kind of forgetting.

The browser opens a page.

The link remains blue.

But the relationship between the citation and what it once referred to has disappeared.

The address survived. The memory did not.

Individual links disappear quietly.

Platforms can disappear all at once.

Google shut down the consumer version of Google+ in 2019 and began deleting content from consumer accounts and pages.

A platform that had hosted years of photographs, conversations, communities, and ordinary social activity stopped being the place where that history lived.

A decade earlier, Yahoo announced the closure of GeoCities.

Before the shutdown, the Internet Archive asked users to help identify GeoCities pages worth preserving, even though its crawlers had already been collecting material from the service.

The rescue effort preserved an extraordinary amount of early web culture.

But the need for a rescue operation reveals something important.

A webpage can feel permanent to the person using it.

Its survival may actually depend on a company continuing to host the platform, maintain databases, renew infrastructure, preserve compatibility, and decide that keeping the service alive is still worth the cost.

What feels like memory at the user level can be maintenance at the infrastructure level.

Digital objects often look self-contained.

They rarely are.

For the original experience to remain accessible, several layers may need to survive at once:

Any one of them can fail.

The photograph may survive while the account disappears.

The text may remain while its images were hosted somewhere else.

The page may load while the embedded video is dead.

An archived copy may preserve the HTML while the database behind an interactive feature is gone.

A domain can expire.

A login system can disappear.

The software required to render the experience can become obsolete.

That is the hidden structure behind digital memory.

Permanence is not one object surviving. It is several dependencies continuing to cooperate.

The obvious response to disappearing web history is archiving.

And without archives, vastly more of the internet would already be inaccessible.

But an archive is not a time machine.

The Wayback Machine itself explains that it cannot preserve every kind of page. Password-protected pages, content available only after submitting forms, and some other types of material may not be captured.

Modern websites make the problem harder.

They are not simply collections of text and images.

Feeds change.

Comments load dynamically.

Maps move.

Videos pull data from other services.

Recommendations depend on databases.

Menus react.

Accounts behave differently depending on who is logged in.

The Library of Congress has documented the challenges that JavaScript and dynamically generated content create for web preservation.

Preserving what a page looked like is one problem.

Preserving the system of interactions that made the page behave like itself is another.

Archives can save extraordinary amounts of history.

But sometimes what survives is a fragment of an environment rather than the environment itself.

Screenshots solve one part of the problem by removing dependencies.

A screenshot can outlive the post it captured.

It can preserve a deleted message, a vanished interface, or a moment from a platform that no longer exists.

Its strength is also its weakness.

It survives by flattening the thing it preserves.

The United States National Archives does not accept static screenshots as substitutes for permanent web records because they do not retain hypertext functionality.

That bureaucratic distinction contains a much larger idea.

A screenshot can preserve what a page looked like.

It does not necessarily preserve:

where the links went,

what happened when you clicked,

which comments surrounded the post,

what appeared before or after it,

how the page changed,

or which other systems made it function.

It preserves the surface.

Not necessarily the relationships.

And relationships are often where context lives.

This is where the old warning about the permanent internet becomes incomplete.

The internet is neither a perfect memory nor a machine that simply forgets.

It remembers unevenly.

Some information becomes extremely durable because copies spread across devices, screenshots, caches, archives, reposts, and other platforms.

Other information disappears because a domain expires, a company shuts down, a URL changes, a database is removed, or nobody thought the material was worth preserving.

Sometimes the trivial object survives.

The context around it does not.

A joke survives without the conversation.

A photograph survives without the community.

A quote survives without the page that explained it.

A screenshot survives without the system that produced it.

The internet can preserve evidence while losing meaning.

Physical archives make maintenance visible.

Books require buildings.

Film requires preservation.

Documents require storage.

Museums require staff.

The web hides much of the same work behind a screen.

When a page opens instantly, the maintenance keeping it available disappears from view.

Servers run.

Domains are renewed.

Software is updated.

Databases are migrated.

Formats remain readable.

Storage is paid for.

Archivists crawl pages.

Someone keeps the system alive.

Digital permanence can therefore feel like a natural property of information.

It is often closer to a continuing agreement:

someone, somewhere, is still maintaining the conditions that allow this thing to remain accessible.

When those conditions disappear, memory can disappear with them.

The old warning was not exactly wrong.

The internet really can remember things for a very long time.

But that is only half the story.

The internet does not forget evenly.

It can preserve the embarrassing moment—

and lose the world that made it understandable.

The Half-Life of a Web Reference — Koehler, Information Research. A longitudinal study of 4,387 URLs examining how web references decay over time.

Perma: Scoping and Addressing the Problem of Link and Reference Rot in Legal Citations — Zittrain, Albert & Lessig, Harvard Law Review. Research distinguishing broken links from references that still resolve but no longer show the originally cited material.

Google+ Consumer Shutdown — Google’s documentation on the 2019 shutdown and deletion of consumer Google+ content.

GeoCities Preserved — Internet Archive’s account of efforts to preserve GeoCities before and after its closure.

Wayback Machine General Information — Internet Archive documentation on what the Wayback Machine can and cannot capture.

Guidance on Managing Web Records — U.S. National Archives guidance explaining why static screenshots do not preserve the functionality of complete web records.