How Long Does a Placement Stay Findable?

What actually happens to a webpage over a decade?

It disappears more often than most people assume. In 2024 the Pew Research Center published a study of digital decay that sampled roughly 90,000 pages per year from the Common Crawl archive for each year from 2013 to 2023 — 999,989 pages in total — and checked in October 2023 whether each was still reachable.

The headline finding is that 38% of webpages that existed in 2013 were no longer accessible a decade later. For a buyer whose interest in a placement is that it persists as a record, that is the single most important number in this series.

Pew defined inaccessible precisely rather than impressionistically. A page counted as gone if it returned a status code of 204, 400, 404, 410, 500, 501, 502, 503 or 523, or failed to return a valid status code at all. This is a conservative test: a page that still loads but has been rewritten, stripped of its content, or moved behind a paywall counts as accessible.

The decay also comes in two distinct forms. Of the 25% of all sampled pages from 2013 to 2023 that had vanished, 16% were individually inaccessible on a root domain that still worked, while 9% were gone because the entire domain had ceased to function. The second kind takes every page on the site with it, which is the failure mode that erases a publication's whole archive at once.

How should that curve be read?

Carefully, because it is easy to misread as good news. The figures fall steadily from 38% for 2013 to 8% for 2023, and the tempting conclusion is that the modern web is becoming more durable.

It is not evidence of that. Every cohort was checked on the same date, so the 2013 pages had ten years in which to disappear and the 2023 pages had a matter of months. The curve is a picture of elapsed time, not of improving practice. The right way to read it is as a decay schedule: this is roughly what happens to a page as it ages.

Read that way, it says something uncomfortable about coverage bought as a durable asset. A placement published today has perhaps a one-in-twelve chance of being unreachable within its first year, and on the experience of the 2013 cohort, better than one chance in three of being gone in a decade.

The non-monotonic middle of the series — 30% for 2016, then 31% and 32% for 2017 and 2018 — is a useful corrective to over-reading. Decay is not a smooth function of age. Individual years are shaped by which sites happened to shut down, and a single publisher's collapse removes a great many pages at once.

Do the links pointing to coverage survive?

Often not, and this matters as much as whether the page itself persists, because a placement that nothing points to is findable only by someone who already knows it exists.

Pew examined links on 500,000 news webpages drawn from 2,063 news sites, sampling half the links on each page for a total of 7,089,514 links analysed. Some 94% of those pages carried at least one external link, with a median of 20 links per page.

Of all outbound links on news pages, 5% were already broken, and 23% of the sampled pages contained at least one broken link. On government sites the figure was similar at 21%, and 39% of external links on government pages redirected somewhere other than their original target.

The Wikipedia figures are the most pointed, because Wikipedia is where sourcing is supposed to be preserved deliberately. Across a sample of 50,000 English-language pages, 11% of all reference links were no longer accessible, 53% of pages containing references had at least one broken link, and on about 2% of those pages every single reference was dead. Coverage cited as evidence decays along with everything else.

Are prestigious outlets safer?

On this evidence, no — and that finding deserves more attention than it gets, because outlet prestige is the organising principle of placement pricing.

Pew sorted news sites into quintiles by traffic and compared them. Among pages on news websites in the top 20% by site traffic, 25% contained at least one broken link. Among sites in the bottom 20%, the figure was 26%. Pew's own summary is that broken links are “about as prevalent on the most-trafficked news websites as they are on the least-trafficked sites”.

That is close enough to identical to treat as no difference at all. Whatever a large news organisation's resources buy, they do not buy measurably better link hygiene than a small one.

A caveat is worth stating plainly: this measures the outbound links on a site's pages, not whether the site's own pages survive. Pew's page-level decay analysis was not broken out by publisher prestige, so the natural follow-up question — do articles on major outlets outlive articles on minor ones — is not answered by this study, and we found no reliable published series answering it. The prestige premium may buy durability; this evidence neither shows that it does nor that it does not.

What does this mean for coverage bought as evidence?

It means the asset has a half-life, and that the systems which examine coverage later will encounter it after the decay has begun.

An immigration petition assembled over several years, a Wikipedia article whose sourcing is challenged a decade after creation, a due-diligence check before an acquisition — each of these reads coverage at a distance from publication. On Pew's figures, a meaningful share of what was purchased will not resolve by then, and USCIS requires that submitted published material carry its title, date and author, which is precisely the metadata that vanishes with the page.

This also reframes what a screenshot is worth. A screenshot proves what a buyer saw; it is not the article, and it is not independently verifiable by a third party later. The durable form of a placement is an archived copy with a resolvable address, not an image in a slide deck.

The practical response is unglamorous and effective: archive at publication. Public web archives accept submissions, capture a timestamped copy, and produce a stable URL that survives the original going dark. Doing that on the day of publication costs nothing and preserves the one property — verifiability by someone else, later — that gives coverage its evidentiary value.

What this means for anyone buying placement

Treat permanence as a property to be secured rather than assumed. Nothing in a placement fee buys durability, and the outlet's size does not appear to buy it either.

Archive every placement on the day it appears, and keep the archived URL alongside the original. If coverage is being bought for an evidentiary purpose — a petition, a notability claim, a regulatory file — the archived copy is the version that will still be working when it is needed.

Record the metadata separately as well: title, publication, date, author, and the original address. These are the elements adjudicators ask for, they are trivial to capture at publication, and they are effectively unrecoverable once a domain has gone.

And weigh durability when choosing where to place. A publication with a long, stable archive and a habit of keeping old material reachable is offering something a higher-traffic site with a history of redesigns and purges may not. Media Morpheus sells placement, and a placement is a page on someone else's website, subject to their decisions and their hosting for as long as they choose to keep it. That is the honest description of the asset, and it is why the archived copy is the part worth holding onto.

Methodology and limitations

All quantitative findings in this report are from the Pew Research Center study When Online Content Disappears, published May 2024. Its page-decay analysis sampled approximately 90,000 pages per year from the Common Crawl repository for 2013 through 2023 (n=999,989) and tested accessibility in October 2023, defining a page as inaccessible if it returned status code 204, 400, 404, 410, 500, 501, 502, 503 or 523, or no valid status code. Its link analysis covered 500,000 news webpages from 2,063 news sites identified using comScore data and collected from the March/April 2023 Common Crawl, sampling 50% of links for 7,089,514 links analysed, plus a sample of 50,000 English-language Wikipedia pages. Five limits apply. First, accessibility is a low bar: a page that loads but has been substantially altered, truncated or paywalled counts as accessible, so these figures understate loss of the original content. Second, the year-by-year series compares cohorts of different ages tested on one date and is a decay schedule, not evidence about changing durability. Third, internal links on news sites were not collected or checked. Fourth, page-level decay was not broken out by publisher prestige, so no finding here establishes whether major outlets preserve their own articles better than minor ones. Fifth, the data were collected in 2023 and describe the web as it was then.

Conclusion

Coverage is routinely bought as a permanent asset and is nothing of the kind. More than a third of the pages that existed in 2013 were unreachable a decade later, a quarter of everything sampled across eleven years was already gone, and nearly one in ten losses took an entire domain down with it. Links fare no better: 5% of outbound links on news pages were broken, more than half of Wikipedia pages with references had at least one dead citation, and the most-trafficked news sites performed no better than the least. None of this is a reason not to buy placement. It is a reason to understand that what is purchased is a page on someone else's server, held at their discretion, and that the only version certain to exist when it matters is the archived copy made on the day it was published.

Frequently Asked Questions

Sources

1. peer-reviewed Pew Research Center, May 2024. When Online Content Disappears: Link Rot and Digital Decay on Government, News and Other Webpages. Authors: Athena Chapekis, Samuel Bestvater, Emma Remy and Gonzalo Rivero.
2. official U.S. Citizenship and Immigration Services. Policy Manual, Volume 6, Part F, Chapter 2, cited for the requirement that published material include title, date and author.