How long do backlinks last? Link rot and survival rates

Most backlinks do not last for ever. Ahrefs found that about two thirds of links had disappeared over roughly nine years, and Linkody's 2026 data shows 63% of monitored links about one year old were still live. Pages get deleted, sites close and links get edited out.

In short

  • An Ahrefs study dated February 2024 found that 66.5% of links pointing to 2,062,173 sampled websites had rotted since January 2013.
  • Linkody's June 2026 analysis of more than 865,000 links monitored by its customers found that 89% of links about three months old and 63% of links about one year old were live when checked.
  • The largest single cause in the Ahrefs study was the linking page being dropped, at 47.7% of lost links, followed by the link being removed from a live page at 34.2%.
  • In the Linkody data, links on domains with a Domain Authority under 10 survived at 33%, against 70% on domains at 30 to 49.
  • Marketplace guarantees vary widely, from three months of removal protection at Collaborator to 12 months at WhitePress, or 36 with its paid warranty, by the vendors' own statements in October 2026.

A backlink lasts as long as three things hold: the page it sits on stays online, the link stays on that page, and the page you were linked to still exists. All three fail over time. The best long-run measurement comes from Ahrefs, which reported in a study dated February 2024 that 66.5% of the links pointing to more than two million sampled websites had rotted since January 2013. Over shorter periods, the only large recent dataset we could open is from the monitoring tool Linkody. In June 2026 it reported that 63% of the tracked links that were about one year old were live when checked. Both figures come with caveats, set out below, and the caveats matter because the two studies measure different populations of links over different lengths of time.

What the studies say

Study Sample Finding
Ahrefs, February 2024 Links to 2,062,173 websites, crawl data from January 2013 66.5% of links had rotted. Counting crawl errors and other problems, Ahrefs put the total lost at 74.5%
Pew Research Center, May 2024 Webpages collected from 2013 to 2023 38% of webpages that existed in 2013 were no longer available. 23% of news webpages contained at least one broken link
Linkody, June 2026 More than 865,000 backlinks monitored by its customers Share of links of each age that were live when checked: 89% at three months, 63% at one year, 46% at two to five years, 31% beyond five years

The Ahrefs and Pew studies describe the web as a whole. Pew’s is not about SEO at all, which makes it a useful independent check: pages vanish at a steady rate whether or not anyone is building links. That changes how link loss should be understood. A good part of it is the ordinary churn of the web, in which sites close and pages are deleted for reasons that have nothing to do with the links on them, so losing links is the normal condition of any profile and not a sign that something was done badly.

The Linkody data is different in kind. It covers links that customers of a monitoring tool chose to track, which are mostly links somebody built or bought. Linkody is a vendor describing its own data, the sample is not random, its analysis excluded trial accounts and competitor tracking, and the age figures are a snapshot of links of each age, not one group of links followed over time. Treat it as a reported picture of built links, not as a law. The snapshot point is easy to pass over and it limits what the figures can say. The 89% and the 63% describe two different sets of links that happened to be those ages on the day of checking, so the difference between them cannot be read as the share of one set that died between month three and month twelve.

Within that picture, Linkody reports an average lifespan of about 20 months for links that die, with 52% of the losses happening inside the first twelve months. If that pattern holds for your own links, the first year is where most of the watching needs to be done, since a little over half of the links that will eventually be lost, 52% on Linkody’s figure, are gone by the end of it. It is also when the arrangement with the publisher is most recent, so early losses are the natural ones to chase first.

The Ahrefs study broke lost links down by cause.

Cause Share of lost links
Linking page dropped from Ahrefs’ index 47.7%
Link removed from a page that still exists 34.2%
Crawl errors 6.45%
301 or 302 redirect 5.99%
Page not found (404) 4.11%
Not canonical 0.82%
Noindex 0.73%

Two causes account for more than four in five losses. Either the whole page dropped out, or an editor took the link out while the page stayed. The two call for different responses. When the page has gone, there is little to ask for except a link from somewhere else on the site. When the page is still there and only the link has been removed, somebody made a decision that can be asked about, and that second group, 34.2% of losses in the Ahrefs breakdown, is where a polite email has something to work with.

Linkody’s 2026 data points the same way. Of all the links it monitored, 56% were live and correct, 17% had been removed from a page that was still live, 14% sat on pages returning errors and 13% hit crawl errors. Under 1% had a changed anchor or attribute. Not every loss is at the far end. If you delete or move the page that was linked to and leave no redirect, the link still exists but points at nothing. That half of the problem is in your hands, and it is the cheapest to prevent, because a redirect added at the time a page is moved saves every link that pointed to the old address without a single message to another site.

What predicts survival

The quality of the site is the factor with a number attached. In Linkody’s data, links on domains with a Moz Domain Authority under 10 survived at 33%, against 70% on domains at 30 to 49. Weak sites are abandoned, sold or cleaned out more often. On those two figures the survival rate on the stronger group of sites is about double that on the weakest, which bears directly on price. A cheap placement on a weak site that is gone within the year can cost more per month of life than a dearer one on a site that is still publishing, though the Linkody figures come from a vendor’s customer data and should be taken as an indication.

The second factor is whether the publisher cares about the page. A link inside an article the publisher wrote for its own readers is tied to content it wants to keep. Practitioners report that niche edits are removed more often than guest posts, because the publisher has invested nothing in the change. That is a reported pattern, not a measured rate. The reasoning behind it is at least plausible: an inserted link is one line that can be taken out in a routine update without anyone missing it, whereas removing a whole article leaves a hole in the site.

Where the page sits also matters. Pages reachable from a category or the homepage tend to stay maintained. Pages buried in an archive are the first to go in a site redesign, because a redesign is the moment a publisher decides which old content is worth carrying over. The last factor is whether the link was paid for. A link placed for a fee is kept for as long as the agreement says, or as long as the seller remembers. A link an editor chose freely has no end date, but it also has no guarantee. Neither kind is safer in every respect, since the paid link comes with terms you can enforce for a period and the earned link comes with no terms and nobody counting the months.

We found no independent study that tracks paid placements separately, so there is no reliable survival rate to quote. What exists are the terms sellers publish. WhitePress’s advertiser terms, read in October 2026, give at least 12 months, with an optional paid warranty that extends this to 36 months. Collaborator stated three months of free removal protection. These are the vendors’ own statements, recorded in our WhitePress desk review and Collaborator desk review. The distance between those two terms shows how much the word “guarantee” can cover, and it is worth setting the stated period against Linkody’s report that links which die last about 20 months on average. A guarantee that ends well before that point covers only the early part of the period in which losses occur.

A guarantee is only useful if you can prove the loss. Sellers also change links without removing them: a followed link becomes nofollow, the page is set to noindex, or the article is moved where no internal link reaches it. In each of those cases the link is still on the page, so a seller can say truthfully that it was never removed, while the reason for buying it has gone. Agree in writing how long the link stays, which attribute it carries and what happens if it changes. Our lesson on buying backlinks with less risk lists the terms to ask for.

The first job is to monitor. Check links you built weekly for the first three months, then monthly. How to track backlinks gives the checks and cadence. The second is to reclaim, because many losses can be reversed. If your page moved, add a redirect. If an editor dropped the link in a rewrite, ask. If the page was deleted, offer a replacement page to link from. Link reclamation is usually the cheapest link work available, because the decision to link was already made once. The third is to claim. If a paid link fails inside its guarantee, claim the replacement or refund promptly and keep the screenshot from the day it went live, since that screenshot is the proof the previous section said a guarantee depends on.

Then budget for it. Treat a share of this year’s links as gone by next year. The share depends on your sources, so measure your own: count links live at 90 and 365 days as a percentage of links built. If you have no history, Linkody’s 63% at one year is a cautious starting assumption for built links, not a forecast. Planning this way changes the target. A team that needs a certain number of live links at the end of the year has to build more than that number, and how many more is something only its own survival figures can say with any accuracy.

Finally, report it. A report that shows only new links hides the loss, and a client reading it will assume the total keeps growing by the number delivered each month. The lesson on link building reports explains how to show survival alongside new links. Showing both is uncomfortable at first, because the net figure is smaller than the gross one, but it is the only version of the report that matches what the client will see when they open a backlink tool themselves.

What we do not know

No public study we could find separates earned editorial links from paid ones with a random sample, and none measures how often a lost link leads to a lost ranking. A link can also stay live and stop counting, if Google drops the page from its index, and no outside tool can confirm that for certain. These gaps set a limit on everything above. The studies say how often links disappear, and they cannot say which of the disappearances mattered, so a survival rate is a measure of what you still hold and not of what it is doing for you.

The sensible response is to spend the effort where the loss is both likely to matter and possible to fix. A lost link from a strong page to an important page on your site deserves a message the same week, while a lost link from a site that has closed deserves a line in the log and nothing more. Fixing broken inbound links is one of the first jobs in a backlink audit, because those are the losses that sit on your own side and can be repaired without asking anyone. Terms such as link rot and link reclamation are defined in the glossary.

Common questions

Are backlinks permanent?

No. A link lasts only as long as the page it sits on stays online, stays indexed and keeps the link. Every large study finds steady losses year after year.

How long does a paid backlink last?

It depends on the terms and the site. We found no independent study of paid placements alone. Guarantees range from a few months to three years, and after that the publisher may remove or change the link.

What is link rot?

Link rot is when a link stops pointing to its intended page, usually because the linking page or the target page was deleted or moved. Ahrefs and the Pew Research Center have both measured it.

Why do backlinks disappear?

The most common reasons are the linking page being deleted or dropped, an editor removing the link during an update, the site closing, a redirect, or the page being set to noindex.

Do lost backlinks hurt rankings?

They can, if the lost links were ones Google counted. One lost link rarely matters. A steady loss that outpaces new links, or the loss of a few strong links to an important page, can show up in rankings.

Vendors to look at

  • BazoomEditor's pick

    Sponsored content and link marketplace, managed service

  • MotherlinkEditor's pick

    Backlink services, guest posts, niche edits, full SEO