Are the links that underpin our digital world quietly disappearing? A team at Harvard University is working to find out, systematically archiving web links even those pointing to obscure, long-forgotten corners of the internet like a 1998 New York Times article referencing a now-defunct Ohio zoning blog. Their work highlights a growing problem: the phenomenon of “link rot,” where once-valid web addresses lead only to error messages.
This quiet, relentless audit has been underway for years. The researchers behind it John Bowers, Clare Stanton, and Jonathan Zittrain of Harvard Law School have built one of the most comprehensive studies of journalistic link decay ever attempted, systematically measuring how the hyperlinks that hold digital journalism together are slowly, invisibly crumbling.
Their findings are striking. Of the 2,283,445 hyperlinks pointing to content outside nytimes.com in articles published between the launch of the Times website in 1996 and mid-2019, 72 percent had rotted meaning the content they once referenced had either disappeared entirely or moved to a new address without redirect. That research, published through the Columbia Journalism Review, represents more than a technical curiosity. It is a quiet alarm about the fragility of digital memory.
Why Links Die
The term link rot also called link death, link breaking, or reference rot describes the tendency of hyperlinks to cease pointing to their originally targeted file, web page, or server over time. The resource may have been relocated to a new address, the host server may have gone offline, or the content may simply have been deleted. A 404 error is the most familiar result, though link rot also includes the subtler problem of content drift, where the URL technically still works but the material behind it has been changed, retracted, or quietly updated without announcement.
The phenomenon has been recognized since the 1990s, when the World Wide Web rose to prominence. Joseph Reagle, associate professor of communication studies at Northeastern University, worked alongside Tim Berners-Lee widely credited with inventing the web at the World Wide Web Consortium during that formative decade. In a 2024 interview with Northeastern's news office, Reagle explained the structural problem at the heart of the issue: URLs serve as address points for web pages, much like physical addresses for homes or offices. The address is powerful because it allows people to find websites easily. The problem is that addresses can be broken, reassigned, or abandoned with little signal or accountability.
The consequences extend far beyond inconvenience. Information professionals have warned that link rot could make important archival data disappear, potentially impacting the legal system and scholarship. A 2014 study by researchers including some on the Harvard team found that nearly half of all hyperlinks in Supreme Court opinions led to content that had either changed since its original publication or disappeared from the internet entirely. The courts, which rely on cited precedent and referenced evidence, are operating with an increasingly incomplete paper trail.
Measuring the Decay
The Harvard team's methodology for the New York Times study was systematic and methodical. Software engineers at the Times extracted URLs embedded in archival articles and packaged them with basic metadata section and publication date. The researchers then wrote a script to visit each unique deep URL and log HTTP response codes, redirects, and server timeouts. Based on this analysis, each link was labeled as either rotted removed or unreachable or intact, returning a valid page.
The scale of the data set is notable: 553,693 articles, more than 2.2 million hyperlinks examined. The 72 percent rotted figure applies specifically to links pointing outside the nytimes.com domain. Internal links, preserved within the Times' own infrastructure, fare better. But the moment a journalist reaches outward to a government report, a blog post, a research paper, a community forum the odds of that reference surviving intact over time drop dramatically.
The team chose to examine the New York Times specifically because of its stature in digital journalism. As the researchers noted, the Times is a well-resourced standard-bearer with a robust institutional archiving structure. That they found such extensive decay suggests the problem is not a failure of individual newsroom effort but a structural feature of how the web operates.
The Scope Beyond Journalism
The Harvard team's work dovetails with broader research into link rot across the internet. Pew Research Center, in a 2023 study, provided additional data points that contextualize the scope of the problem. According to Pew's findings, a third of all web pages that existed at one point between 2013 and 2023 are no longer accessible. Among news web pages specifically, 23 percent contain at least one broken link. Government websites show a similar rate, at 21 percent. Wikipedia, one of the most carefully maintained and widely cited resources on the internet, is not immune: 54 percent of Wikipedia pages contain at least one link in their References section that points to a page that no longer exists.
The rot is not evenly distributed. A 2003 study found that roughly one link out of every 200 broke each week on the web, suggesting a half-life of 138 weeks approximately two and a half years. A 2016-2017 study of links in the Yahoo! Directory, which had stopped updating in 2014 after 21 years of development, confirmed this rate broadly, finding the directory's links had a half-life of approximately two years. Studies within digital libraries show slower decay roughly 3 percent of objects became inaccessible after one year in one 2002 analysis because institutional archives tend to maintain their collections with more care than commercial web hosts.
Even social media is not immune. Nearly one-in-five tweets are no longer publicly visible on the site just months after being posted, according to Pew's data. The ephemerality that users often associate with platforms like Twitter or Instagram is not merely a perception it is a documented property of the infrastructure.
A Tool for the Problem
Researchers and archivists have not waited passively for the problem to worsen. One response has been Perma.cc, a service designed to prevent link rot by capturing web content at the moment a citation is made and preserving it in a persistent archive. In July 2026, Ohio University Libraries announced it was adding access to Perma.cc for its community, joining a growing number of academic institutions that have integrated the tool into their research support services.
According to Ohio University's announcement, Perma.cc provides a more thorough and accurate capture than the Internet Archive's Wayback Machine, along with persistent shortlinks that are more convenient for citation and the ability to manage links with folders, annotations, and visibility controls. For sources that lack a DOI the digital object identifier that serves as the gold standard for citing scholarly materials Perma.cc offers a practical alternative that preserves the web as it existed at a specific moment in time.
The service fills a gap that Digital Object Identifiers cannot cover. DOIs are generally available only for scholarly sources like journal articles, book chapters, and government reports. The vast universe of web-based content blog posts, news articles, data sets, community forums has no equivalent persistent identifier. Perma.cc attempts to create that stability retroactively, at the moment a researcher or journalist decides a link is worth preserving.
What the Research Reveals About Digital Memory
The Harvard team's framing of the problem distinguishes between link rot and content drift. Link rot is the dramatic failure a page gone, a URL dead. Content drift is more insidious: the page still exists, but the content has changed. A news article is rewritten. A research paper is retracted. A corporate statement is edited to remove language that was once cited as evidence. The URL persists, but the record it points to is no longer the record that was originally referenced.
This distinction matters for journalism specifically. A news article is often a record of what was known and said at a particular moment. When the linked sources that informed that record disappear or change, the article becomes a shell of its former self unable to substantiate its claims, unable to provide readers with the context that was once one click away. The hyperlink, which journalists use to provide depth and evidence, becomes an artifact of a web that no longer exists.
The research team at Harvard has been clear that their goal is not to assign blame or to criticize any individual newsroom. The problem is structural, baked into the architecture of the web and the economics of hosting. When a company goes bankrupt, a blog is abandoned, or a government agency restructures its website, the URLs that pointed to that content often go with them. There is no guarantee that a link made in good faith in 2005 will still work in 2025 or in 2026, as the case may be.
Why This Matters for PostsNews Readers
For readers researching practitioners, frameworks, and ideas particularly those who rely on linked sources to support their own work the implications are practical and immediate. The next time you follow a link in an article about media research, there is roughly a one-in-four chance that link leads somewhere broken. For older content, the odds are worse. A news article from 2015 that links to a source cited as evidence may be pointing to a page that no longer exists.
This is not a reason to distrust linked sources, but it is a reason to verify them. When researching a claim or a framework, the existence of a link is not the same as the existence of the content the link pointed to when it was written. Checking whether a cited source is still accessible and, if possible, using tools like the Wayback Machine or Perma.cc to find archived versions is now a basic skill for anyone working with digital research materials.
The Harvard team's work suggests that the problem is getting worse over time, not better. As the web expands, the volume of potentially rotted links grows. The half-life of a link, measured in years more than decades, means that most hyperlinks have a finite lifespan. Building systems that acknowledge this fragility that capture content when it is alive, that use persistent identifiers where they exist, that document the state of a source at the moment of citation is increasingly essential for any field that relies on written records.
The Researchers and Their Approach
John Bowers, Clare Stanton, and Jonathan Zittrain bring different backgrounds to the problem. Zittrain, a professor at Harvard Law School, is well known in internet law and policy circles. Bowers and Stanton contributed empirical rigor and technical skill to the project. Together, they have produced research that is both a diagnostic tool measuring the extent of a problem and a provocation, inviting the field to think more carefully about what it means to rely on hyperlinks as a form of citation and evidence.
The team's work has not produced a solution that eliminates link rot no such solution exists, given the distributed and economically motivated nature of web hosting. But it has produced something valuable: a clear-eyed accounting of the scope of decay, a vocabulary for distinguishing different types of failure, and a reminder that the permanence of digital information is an assumption, not a guarantee.
For Reagle, who has thought about these issues since his days at the World Wide Web Consortium in the 1990s, the problem requires infrastructure changes and human collaboration to solve. URLs are addresses, and addresses can be abandoned. The web has no mandatory archiving requirement, no mechanism that ensures every published document is preserved indefinitely. What exists instead is a patchwork of archival efforts the Internet Archive's Wayback Machine, institutional repositories, services like Perma.cc that attempt to catch what would otherwise be lost.
Where the Field Stands
The research on link rot continues. The Pew data from 2023 provides a broad snapshot of current decay rates across different types of websites. The Harvard team's work offers a deep dive into one of the most important publications in American journalism. The work at Ohio University and other institutions to deploy Perma.cc reflects a growing recognition that the problem is not merely technical but archival that the citation practices of scholarship and journalism need tools that assume impermanence more than permanence.
The researchers are careful not to be alarmist. Link rot is a feature of a living, changing web, not a bug. Content moves, servers go offline, organizations restructure. The web is not a library; it is a distributed, economically motivated system for publishing information that was never designed with indefinite preservation in mind. The challenge is not to eliminate change but to build systems that account for it that allow readers and researchers to understand what a source looked like when it was cited, not just whether the URL still resolves.
That is the work that continues, quietly, server by server, link by link, in Cambridge and beyond.
Where to Read Further
For readers interested in the full Harvard study and its methodology, the Columbia Journalism Review analysis of the New York Times link rot data is available at CJR's analysis of linkrot and content drift. The broader research on link rot prevalence, including the Pew Research Center data on web page accessibility, is covered in Northeastern University's report on link rot and internet infrastructure. For practitioners seeking tools to prevent link rot in their own work, Ohio University Libraries' guide to Perma.cc provides practical information about account access and use.