Link Decay: Why Backlinks Disappear and What to Do
•September 17, 2026•9 min read
A backlink is not a permanent asset. It is a line of HTML on a page someone else controls, and pages get edited, redirected, deleted and abandoned every day. Link decay, sometimes called link rot, is the slow loss of links you already earned or paid for. It is invisible in the month it happens and very visible a year later, when a page that used to rank has quietly lost a third of the domains pointing at it. This article puts numbers on the problem, explains why links die, and lays out a maintenance routine that keeps a profile from shrinking under you.
TL;DR. Ahrefs tracked links to 2 million sites over nine years and found 66.5% of them gone. Compounded, that is roughly one link in nine lost every year, before you build anything new. Almost half of the losses are pages that simply disappeared; a third are links an editor removed. Links from news sites and dated posts rot fastest; links inside evergreen content that still ranks last longest. Losing links hurts: when Ahrefs disavowed links to its own pages, traffic fell 13 to 18% and came back when the file was removed. The fix is a routine, not a rescue: a monthly lost-links check sorted by cause, a small recovery effort where recovery is possible, and a standing replacement budget for the rest, spent on placements built to last.
How fast links actually disappear
The largest study on the subject is Ahrefs' link rot study, which followed the links pointing at 2,062,173 websites from January 2013 onward. Nine years later, 66.5% of those links had rotted; counting temporary errors as well, 74.5% were lost at the time of the check. Spread evenly, 66.5% over nine years is a compounded loss of about 11% a year: a profile of 300 referring domains that earns nothing new is a profile of roughly 265 twelve months later, and around 200 after three years.
The decay is not unique to SEO links. Pew Research Center's 2024 report When Online Content Disappears found that 38% of web pages that existed in 2013 were no longer accessible a decade later, and that 54% of English Wikipedia articles contain at least one reference link that leads nowhere. Age is the strongest predictor. A Harvard study of 2.3 million links in New York Times articles, summarised in the Columbia Journalism Review, found that 6% of links published in 2018 had rotted, against 43% of links from 2008 and 72% of links from 1998. A link is most likely to survive its first year and least likely to survive its tenth.
Why links die: the five causes
Ahrefs broke its lost links down by reason, and the split matters because each cause calls for a different response:
| Cause | Share of lost links | What it means for you |
|---|---|---|
| Page dropped | 47.7% | The linking page, and often the whole site, is gone from the index. Nothing to recover; replace the link. |
| Link removed | 34.2% | The page is alive but an editor took the link out: a refresh, a pruning of external links, or an expired paid placement. Sometimes recoverable by asking. |
| Crawl error | 6.45% | The page could not be fetched at check time. Often temporary; re-check before acting. |
| Page redirected | 5.99% | The linking URL now redirects elsewhere and the new page may not carry the link. Ask for the link on the new page. |
| Page not found | 4.11% | A 404 on a site that still exists. The page may return; usually it does not. |
Read as a whole, the table says something uncomfortable: more than half of all link loss (dropped pages, not-found pages, redirects) is out of your hands, because it comes from the linking site's own decisions about its content. Only the "removed" third is ever negotiable, and much of that is the natural end of rented placements, sponsored posts with a twelve-month term, and directory listings that expire when the subscription does.
Which links rot fastest
Not all links decay at the same rate, and the pattern is consistent across the studies.
- News and dated content rots first. Pew found that 23% of news web pages contain at least one broken link, and the Harvard study showed news links losing accessibility steadily with age. News sites archive, paywall, restructure and delete more than any other category, and a link in a story that stops being read stops being crawled, which is its own slow death.
- Links in content that no longer ranks fade even while they exist. The Harvard team also measured "content drift": 13% of still-reachable links had changed meaning since publication. A backlink inside a page that has been rewritten three times, or that has slipped out of the index, is technically alive and practically worthless.
- Rented placements have an expiry date by design. Sponsored posts sold on annual terms, link insertions with a renewal fee, and directory listings are the part of the "removed" 34.2% you can predict to the month.
- Links from sites without real traffic die with the site. The 47.7% of losses from dropped pages are dominated by domains that were never viable: expired-domain networks, abandoned blogs, sites that existed to sell links and folded. We showed how to spot them in advance in Real Traffic vs Fake Metrics.
- Evergreen, in-content links on pages that keep ranking last longest. A page that earns organic traffic gets maintained, gets crawled, and gives its owner a reason to keep it live. Its outbound links inherit that stability.
Does losing links actually hurt?
It does, and the cleanest evidence comes from Ahrefs disavowing its own links. In Do Links Still Matter for Rankings?, Patrick Stox disavowed 3,476 links pointing at three Ahrefs blog posts, left the file in place for four weeks, then removed it. One page lost about 18% of its search traffic in the first month; another lost 13.3% and returned to 99% of its previous traffic once the disavow was lifted; a third rose by over 44% after the file came out. A disavow is not identical to decay, but from Google's side the effect is the same: links that used to count stop counting. The difference is that decay never gets reversed by deleting a file. Whatever you lose stays lost until you replace it, which is why a profile that is not maintained ranks a little worse every quarter without any obvious event to blame.
How to catch decay before rankings move
Rankings are a lagging signal; the lost-links report is the leading one. A monthly routine takes under an hour:
- Pull lost links for the last 30 days from your backlink tool and group them by the five causes above. Ignore crawl errors for a month; they usually resolve. Ignore links from domains that had no traffic; they were not helping.
- Weigh the loss by the target page. A lost link to a money page that ranks on page one matters ten times more than a lost link to a blog post from 2021. Track referring domains per key page, not per site.
- Check the linking site, not just the link. If the domain lost its own traffic, its remaining links to you are next. Mark them as at-risk.
- Flag anything you paid for. If a placement with a stated term is approaching renewal, decide now whether it earned its keep; if a placement sold as permanent has vanished, that is a publisher to stop buying from.
- Compare the run rate with your build rate. If you are losing eight domains a month and adding five, the profile is shrinking however good the new five are. Most sites discover this a year late.
What to do about it
Match the action to the cause, and be honest about the recovery rate. Outreach to recover removed links works occasionally; outreach to recover dropped pages never does.
| Cause | Try to recover? | Action |
|---|---|---|
| Page dropped / not found | No | Replace. Add the domain to a do-not-buy list if it sold you the link. |
| Link removed, page alive | Yes, once | One short email: the page still cites your data or product, the link is missing, here is the URL. Then replace if no answer in two weeks. |
| Page redirected | Yes | Ask for the link on the destination page. If the redirect is to a category or the home page, treat as dropped. |
| Paid placement expired | Renew or replace | Renew only if the page still ranks and sends traffic; otherwise move the budget to a placement that does not expire. |
| Crawl error | Wait | Re-check next month before spending anything. |
Then plan the replacements as a standing budget rather than an emergency. At an 11% annual loss, a profile of 300 domains needs about 33 new relevant domains a year just to hold its ground, spread across the year so the additions look like a heartbeat rather than a spike. We covered why the spread matters in Backlink Velocity. The pace is cheap to hold when the replacements are ordered as editorial placements from a vetted marketplace, where donors are filtered by topic and real traffic and the price per link is known in advance.
Buy links that are built to last
Since roughly half of decay comes from the linking site failing, the biggest lever is choosing sites that will still exist and still rank in three years. That means the vetting steps from How to Vet a Publisher: real organic traffic, a topic that matches yours, a publishing history longer than a year, and no sign the site is an expired domain wearing an old brand. It also means preferring the format that survives: a standalone, in-content article on the donor site rather than a link inserted into someone else's old post, which is the first thing removed in a refresh. Placements ordered through Rixot are published as their own article on the donor site with no renewal term, so the link's lifetime is the article's lifetime, and the donor pool is filtered to sites with measurable traffic for exactly the reason the Ahrefs numbers show: dead sites take their links with them. Finally, keep a copy of every placement URL and check it quarterly; a disappeared article on a site that is otherwise healthy is a support ticket, not a loss.
FAQ
What is link decay?
The gradual loss of existing backlinks as linking pages are deleted, redirected, edited or de-indexed. Ahrefs measured 66.5% of links lost over nine years, roughly 11% a year compounded.
Why do backlinks disappear?
Mostly because the linking page drops out of the index (47.7% of losses) or an editor removes the link (34.2%). Redirects, crawl errors and 404s make up the rest. Paid placements with a fixed term expire on schedule.
Does losing backlinks lower rankings?
Yes. Ahrefs' disavow experiment on its own pages produced traffic drops of 13 to 18% while links did not count and a recovery once they did. Decay produces the same drop with no file to remove.
How do I stop link decay?
You cannot stop it, only outpace it: check lost links monthly, recover the few that are recoverable, and replace the rest at a steady rate with in-content placements on sites that have real traffic and no expiry on the link.
