What Happens to Award-Winning Websites Five Years Later

I keep a bookmark folder of sites I admired. I opened it a few weeks ago, which I don’t recommend as an activity.

Maybe a third of the links resolved to something. Another third gave me a 404 or a parked domain with an offer to buy it. And the last third technically loaded, but arrived as a stack of broken layout, missing fonts, and a hero video that spins forever without playing. Those are the worst ones. A dead link at least tells you it’s dead.

A building decays slowly and in public. Everyone can see the paint go. A website works perfectly right up until the Tuesday afternoon it doesn’t, and usually nobody notices for months.

The Web Doesn’t Keep Things

Pew Research put numbers on this in May 2024, in a study I think about more than is healthy. They pulled a random sample of just under a million pages from the Common Crawl archive, one batch for each year from 2013 to 2023, then tried to load every one.

A quarter of everything that existed at any point in that decade was already gone. Of pages that existed in 2013, 38% were unreachable ten years later. Even pages from 2023 — less than a year old when they checked — failed 8% of the time.

Eight percent. In under twelve months.

Their definition was conservative, too. They only counted pages returning an error code, meaning the page or its host server no longer exists. Pages that still load but have quietly become something else were outside the scope. So the real decay rate, if you count “the URL works but the thing you remembered isn’t there,” is worse than 25% and nobody knows by how much.

Ahrefs ran a separate crawl and put the share of dead links at above 66% over nine years. Different method, different sample, same direction.

None of this is specific to award winners. But award winners have a particular problem, which is that they were usually built at the technical edge of what was possible at the time. And the edge moves.

What Actually Kills Them

Flash is the clean case, because it has a death certificate with a date on it.

Adobe announced end of life in July 2017 and pulled support on December 31, 2020, then started blocking Flash content outright on January 12, 2021. That’s three and a half years of notice, which is more warning than most technologies get. It didn’t matter. An enormous amount of the most celebrated interactive work from roughly 1999 to 2012 simply stopped running. Not degraded. Stopped.

Here’s the part that stings if you care about this stuff: the Smithsonian noted that digital archivists had real difficulty capturing and replaying Flash sites with standard web archiving tools. The work wasn’t just taken offline. In many cases it couldn’t be properly saved in the first place. Preservation projects have done heroic work since — Ruffle, an open-source emulator written in Rust, and the Flashpoint Archive, which has rescued over 160,000 pieces of Flash content, and the Internet Archive, which folded Ruffle into its own collection. Newgrounds sponsored Ruffle to keep its own library alive.

But an emulator is a projector for a film format nobody manufactures anymore. It’s a rescue, not a life.

Flash was the loud extinction. The quiet ones are still going.

Frameworks reach end of life. Whatever was current when a site was built has a support horizon, and once it passes, security patches stop and the site becomes a liability on somebody’s risk register. That’s usually how the decision to kill it gets made — not by a designer, by whoever signs off on the audit.

The CMS gets abandoned. A bespoke backend built for one project is fine while the agency that made it still exists and still remembers how it works. Five years is long enough for both of those to stop being true.

Third parties vanish. The font service, the video host, the map embed, the analytics script. Each one is a dependency you don’t control, and each one has its own end-of-life date that nobody told you about.

Somebody forgets to renew the domain. Genuinely, this is a leading cause of death for beautiful things.

Or the client just rebrands. Which is the most common outcome and the least dramatic. The work was excellent, the company pivoted, the URL now redirects to a template.

What Actually Survives

I’ve noticed something looking through old winners, and I don’t fully trust it as a rule, but it holds up more often than not.

The sites that still work five and ten years on tend not to be the ones that were technically ambitious. They’re the ones that were structurally simple and conceptually stubborn. Real HTML, real text, few dependencies, and one strong idea carried all the way through instead of a stack of effects.

The ones that died young were mostly built on a bet — that a plugin would persist, that a browser behavior would stay supported, that scroll-jacking would still feel clever in 2019. Bets have expiry dates. Decisions don’t.

Which is roughly the difference between what wins attention and what earns it back later. Go look at any collection of awwwards web design work from a few years back and sort it in your head into those two piles. The technique-forward entries have aged badly, sometimes hilariously. The ones built on an actual point of view mostly haven’t aged at all, because a point of view doesn’t have a runtime dependency.

Does that mean ambitious work is a mistake? No, and I’d argue the opposite. Somebody has to push, or the whole medium stops moving. But there’s a difference between pushing on ideas and pushing on infrastructure, and the second kind comes with a shelf life that nobody puts on the invoice.

Building Something That’s Still There in 2031

Assume every dependency will die and decide which deaths you can survive. Self-host the fonts. Keep the critical content in HTML rather than assembling it in JavaScript at runtime. If your homepage needs four external services to render a sentence, it has four separate expiry dates.

Make it degrade instead of collapse. Fancy animation that fails should leave a readable page behind, not a blank screen with a spinner. This is a five-minute decision at build time and it’s the difference between a site that ages and a site that dies.

Archive it on purpose, on launch day. Submit it to the Wayback Machine yourself. Keep a full-resolution video walkthrough, keep the design files somewhere that isn’t one person’s laptop, and write down the stack and the reasoning. Your own case study is a preservation format, and most agencies discover this two years after the client killed the site.

Own the domain longer than the relationship. Auto-renew, ten years, done. It costs less than one hour of design time.

And know what you’re building. A campaign site with a six-week life is a different object from a company’s primary site, and pretending otherwise wastes money in one direction or loses work in the other.

Here’s the thought I can’t shake. The most celebrated web work of the 2000s is now mostly viewable through emulators maintained by volunteers, if it’s viewable at all. Somebody won a prize for it. Somebody presented it at a conference. And the only reason any of it still exists is that a handful of people with no commercial stake decided it was worth saving.

What are you building right now that would survive that test? And who exactly do you imagine is going to bother?