How long does a backlink actually last?

Nobody can tell you the average lifespan of a backlink, and the figures you find quoted come from samples that resemble no one in particular. A citation in a news archive and a link on a young company's homepage have nothing in common except the word "backlink".

What can be said usefully is that links rarely fade. They disappear at a small number of identifiable moments, and those moments can be seen coming more often than people think.

The five moments that kill links

A redesign. The most common by far. The new site keeps the content and loses the outbound links, or moves them into a template that no longer renders them. Nobody involved thinks of it as a decision. If the redesign is yours, the problem is entirely different and the check-list is here.

A CMS migration. Content is exported, links inside the body survive, links that lived in a sidebar module do not. Six months later the old module is gone and so is your placement.

An editorial clean-up. Someone is asked to reduce outbound links, or to review "old articles". Your link is not targeted; it is simply in the sweep. This is the case where a polite message works best, because the person on the other side had no particular reason to remove it.

A change of hands. The site is sold, the author leaves, the freelancer who wrote the piece is no longer maintaining it. The new owner has no reason to keep a link they did not place, and often no idea why it was there.

The site simply ends. The domain expires, the blog stops. This is the only case where nothing could have been done, and also the rarest.

What published lifespans are worth

You will regularly read that a backlink "lasts two years on average", or that a third of links disappear within a year. Those figures come from samples built on what a tool observes, which is overwhelmingly links obtained through campaigns, on sites that sell visibility. They describe that population accurately, and not yours.

The useful reference point is not an average, it is a distinction: links a site has its own reason to keep survive more or less indefinitely, and the rest leave at the first change. There is no continuum between the two, there are two populations.

Which links stay

The pattern is simple: links a site has its own reason to keep. A reference inside an article that still gets read. An entry in a resource page someone is proud of. A citation in technical documentation, which survives almost anything because removing it would mean checking whether the information changed.

Links that exist only as a favour, a partner exchange, a placement negotiated once, are the first to go, because nobody on the other side is invested in them. That is also one more argument for judging a link by whether it deserves to be there rather than by a domain score: the question "would this page exist without me?" predicts lifespan better than any metric.

The signals that announce a disappearance

Three of them are visible without any particular effort:

• An announced redesign. Many sites mention it, in a post, a banner or a newsletter. That is the moment to write, before the build rather than after.
• An author leaving. The byline changes, the former contributor announces their move elsewhere. The pages they maintained become orphans and are the first to be cleaned up.
• A domain close to expiry. That information is public, and an approaching date with no renewal says something about the attention the site is getting.

None of these calls for immediate action. They only tell you where to look first, which is exactly what a signal is for.

Recovering a link: what works, and when

Recovery depends almost entirely on elapsed time. In the first few weeks the article is still recent, the person who wrote it is still there, and a message pointing at a specific paragraph often gets the link restored without discussion. After six months you are writing to someone who remembers neither the article nor you.

The message that works never mentions SEO. It reports a reading problem: a passage that cites a source and no longer leads anywhere. That is true, it is verifiable, and it gives the person a reason to act that is not doing you a favour.

What dies without disappearing

A link can stop counting while remaining perfectly visible, and those deaths fit none of the five moments above, because they leave no trace at all.

Switching to nofollow is the classic case: the link is there, it is clickable, and an attribute added by a security plugin or a new CMS default has stripped away what made it valuable. What that attribute actually changes is worth knowing before you discover it on your best placement.

A rewritten anchor is quieter still. The editor reworks a paragraph, and "their guide to migration redirects" becomes "here". The link lives, it is followed, and it has lost the description that gave it meaning: anchor text is the shortest description of your page anyone will ever write, and nobody will tell you it changed.

In practice these two cases are more common than outright removal once the second year has passed. A lifespan measured only on the presence of the link counts them as living links, which explains part of the gap between published figures and what you actually observe.

Should you replace a lost link?

Often not, and it is the question people rarely ask. A link that disappeared because the page carrying it no longer interested anyone was not worth much to begin with. Replacing it with an equivalent link amounts to buying the same problem twice.

The loss deserves effort when it concerns a link that sent traffic, that came from a site you would struggle to get today, or that you told someone about. In every other case the right answer is to note the disappearance and move on.

What this changes in practice

The useful question is therefore not "how long will this link last" but "will I notice when it goes". A link removed in a redesign is often recoverable: the site still exists, the relationship still exists.

The same message sent eighteen months later, to someone who has since replaced the whole site, does not work. That gap between the removal and your discovery of it is the only part of this you actually control: which is what monitoring is for.

And if you track all this in a spreadsheet, write the date of the last check next to each link rather than a status: it is the column that lies least, because it states its own age. A link checked yesterday and a link checked in March look identical in any status column, and they are not the same thing at all. The date is the only field in a tracking sheet that cannot quietly become wrong, because it describes when you looked rather than what you saw.