Most backlink advice treats a link as a one-time event. You earn it, it counts, done. That's never been fully true, and it matters a lot more now than it used to. For the fundamentals of how a link earns authority in the first place, see How Do Backlinks Work in SEO?
Backlinks Die at a Predictable Rate
Ahrefs analyzed more than two million domains and found that 66.5% of links pointing to those domains since 2013 were dead by the time of the study (Ahrefs, "At Least 66.5% of Links to Sites in the Last 9 Years Are Dead," via Broken Link Checker's coverage). That works out to roughly 7% of a backlink profile dying every year, whether or not anyone notices.
Links die for ordinary reasons: pages get redesigned, content gets deleted, sites shut down, URLs change without a redirect. None of it is unusual or a sign anything went wrong. It's just what the web does over time, and most backlink tracking treats a link as counted the day it's found, not monitored for whether it's still alive months or years later.
AI Systems Are Built to Prefer Fresh Sources
This is the part that changes the stakes. In traditional SEO, an old page with strong backlinks could hold its ranking for years. In AI search, freshness is now one of the strongest citation signals systems use.
Multiple 2026 studies converge on similar figures: roughly half of all AI-cited content is under 13 weeks old, and content published in the last 30 days earns an estimated 3.2x more AI citations than older pages (Authority Tech, "Content Freshness in 2026"). This isn't a soft best practice. Google's own generative AI optimization guide, published in May 2026, explicitly describes retrieval-augmented generation, the mechanism behind AI Overviews and AI Mode, as retrieving "relevant, up-to-date web pages" from its index (referenced in Authority Tech's coverage).
The pattern shows up consistently across industries: a page cited weekly in March can drop to monthly citations by August and disappear from AI answers entirely by year end, even while its HTML still renders fine and its Google ranking holds steady (Pravin Kumar, "Why AI Engines Stop Citing Your Pages"). The page looks healthy in every traditional metric. It's decaying anyway, just in a layer most reports don't check.
Why This Hits AI Citation Harder Than It Hits Rankings
Google AI Overviews shows the weakest freshness bias of the three major AI systems, citation patterns there still track fairly closely with traditional ranking age, so an established page with strong backlink authority can keep earning AI Overview citations even without a recent refresh (Demand Local, "Content Freshness AI Rankings"). ChatGPT and Perplexity show a much stronger freshness preference, which means the same aging page can be actively falling out of citation rotation on those platforms while it still looks fine in a Google-focused report.
This is exactly why decay is easy to miss. Whoever's watching Google rankings sees stability. Whoever isn't watching AI citation platforms separately doesn't see the erosion happening one layer up.
Even Wikipedia Isn't Immune
If any site should be protected from this, it's the one AI systems cite constantly. It isn't. Pew Research's link rot study found that 54% of Wikipedia pages contain at least one dead reference link in their citations section (Pew Research, via Broken Link Checker).
AI systems increasingly perform a freshness and validity check on cited URLs as part of generating an answer. A dead outbound link on an otherwise strong page gives the system a measurable reason to demote that page in the citation ranking for that query (Broken Link Checker, "How Broken Links Hurt Your AI Search Visibility"). Every backlink pointing to a page that now returns a 404 is authority the AI system can't actually credit, whether or not it still counts in a backlink tool's total.
How to Check Whether Decay Is Already Costing You
A few concrete checks reveal whether this is already happening to a site rather than just being a theoretical risk.
Audit backlink health by age, not just by volume.
This is part of maintaining a healthy backlink profile over time, not a one-time check.
Most backlink tools show total referring domains at a point in time. Pull a historical view instead. In Ahrefs' Site Explorer, use the Referring Domains report and switch the chart to show both gained and lost domains over the trailing 24 months, not just the current total. In Semrush's Backlink Analytics, the equivalent is the "New vs. Lost" backlinks tab. Either way, the number to watch is the lost-domains trendline, not the net total, since a flat or slowly growing net count can hide a large amount of churn underneath it.
Segment lost links by age at time of loss.
A link that died within its first six months usually points to something different (a bad placement, a low-quality site that shut down) than a link that survived two years and then died. Sorting lost links by how old they were when they broke helps distinguish normal attrition from a specific, fixable problem.
Check your own outbound links, not just inbound ones.
A page citing a study or source that's since gone dead is vulnerable the same way Wikipedia is. Run key pages through a broken-link checker (Ahrefs' Site Audit has a built-in broken-links report, or a dedicated tool like Broken Link Checker works for a quick manual pass) and fix or replace any dead outbound citations, especially on content that's meant to be a durable reference.
Compare Google Search Console rankings against AI citation tracking for the same pages.
Pull the Performance report for a target page, filtered to the last 16 months, and look at average position specifically. If position has held steady or improved, but a citation-tracking tool, or a manual spot-check across ChatGPT, Perplexity, and Google AI Mode for the page's target queries, shows citation frequency dropping, that's the exact pattern this article describes: stable in one layer, decaying in the other.
Flag content approaching the one-year mark.
Several independent studies converge on roughly one year as the point where content starts falling out of most AI citation pools entirely, unless it has accumulated enough authority to be treated as a canonical source regardless of age (Demand Local, "Content Freshness AI Rankings"). Pages crossing that threshold are the highest-priority candidates for a substantive refresh, not just a date-stamp update.
What This Looks Like Over Time
Here's an illustrative walkthrough of how this typically plays out, not a specific client's numbers, but a representative pattern worth having in mind while auditing your own site.
A page publishes and earns 40 backlinks over its first six months, along with steady AI citation activity on its target topic. By month 12, 4 to 5 of those links (roughly the 7% annual decay rate) have gone dead, referring domains from redesigned or shut-down sites, mostly invisible in a standard backlink report since the net domain count is still climbing from newer links replacing old ones. Google ranking holds steady through this whole period, since ranking algorithms weight accumulated authority more heavily than pure recency.
AI citation frequency tells a different story. Around month 9 to 10, the page's citation rate on ChatGPT and Perplexity starts declining, even though nothing about the page itself has changed and its Google position hasn't moved. By month 14 to 16, it's largely absent from AI answers for queries it used to appear in regularly, while still ranking on page one of Google. The gap between "still ranking fine" and "no longer being cited" is exactly the blind spot this kind of audit is built to catch, and it typically opens up months before it would show up in any dashboard that only tracks rankings.
Fixing It: Reclamation, Refreshing, and Monitoring
Reclaim what's fixable first, since it's the highest-return fix available.
A dead inbound link is often just a moved page, not a deleted one. Search for the linking site's current version of the page (a quick site search or a Wayback Machine check usually finds it), then send the site owner a short, specific note:
Subject: Quick fix for a dead link on [Page Title]
Hi [Name], I was reading your article on [Topic] and noticed the link to [Your Original URL] returns a 404, it looks like the content moved to [New URL]. Just flagging it in case it's useful, happy to help if there's anything else I can clarify about the resource.
This works because the site owner already made the decision that your content was worth linking to once. You're not asking for a new favor, you're helping them fix something that's quietly hurting their own page too, since a dead outbound link isn't great for them either. Response rates on this kind of outreach run meaningfully higher than cold link-building pitches, since there's no ask involved beyond a courtesy.
Refresh substantively, not cosmetically.
Updating a publish date without changing the actual content doesn't move the freshness signals AI systems are checking. A real refresh adds new data, corrects anything outdated, and genuinely re-earns the "recently updated" status rather than faking it.
Fix your own outbound links before they cost you a citation.
Since AI systems can demote pages with broken outbound references, treating outbound link health as part of routine maintenance, not just inbound link building, protects citation eligibility directly.
Build monitoring into a regular cadence, not a one-time cleanup.
Since roughly 7% of a backlink profile decays annually, a single audit doesn't solve this permanently. Treating link health as ongoing maintenance, the same way uptime or site speed gets monitored, catches decay before it compounds. A quarterly check of the lost-domains trendline and a spot-check of AI citation status on your top 10-15 pages is enough to catch most decay before it meaningfully affects visibility.
When we track backlink health for client sites over time, the pattern holds up consistently: the sites that treat this as continuous monitoring rather than an occasional cleanup are the ones whose AI citation numbers stay stable instead of quietly eroding between check-ins.
A New Tool Worth Knowing About
On February 4, 2026, the Internet Archive and Automattic released the Wayback Machine WordPress Link Fixer plugin, the first widely available, automated tool for detecting and fixing broken outbound links at scale, free and open source, listed in the official WordPress.org plugin directory (Broken Link Checker, "Link Rot in 2026"). For any WordPress site, this turns outbound link maintenance from a manual, easily-neglected task into background infrastructure. It doesn't fix inbound link decay, that still requires outreach and reclamation, but it closes the outbound half of the problem with minimal ongoing effort.
Frequently Asked Questions
How fast do backlinks actually decay?
Ahrefs' analysis of more than two million domains found that 66.5% of links pointing to those domains since 2013 were dead by the time of the study, roughly 7% of a backlink profile decaying per year. This happens through ordinary causes like site redesigns, content removal, and domain changes, not through any error in how the links were originally built.
Does content freshness really affect AI citations?
Yes, more than it affects traditional Google rankings. Multiple 2026 studies found that content under 30 days old earns an estimated 3.2x more AI citations than older content, and roughly half of all AI-cited content is under 13 weeks old. Google's own generative AI optimization guide describes retrieval-augmented generation as pulling "relevant, up-to-date web pages," confirming freshness is a built-in mechanism, not just a correlation.
Why would a page hold its Google ranking but lose AI citations?
Google AI Overviews shows a weaker freshness bias than ChatGPT or Perplexity, so an established page can keep its traditional ranking and its AI Overview citations even without a recent update. The same page can be losing citations on ChatGPT and Perplexity at the same time, since those platforms weight recency more heavily. Tracking rankings alone won't reveal this gap.
Do broken outbound links hurt AI citation eligibility?
Yes. AI systems increasingly run a freshness and validity check on cited URLs as part of generating an answer, and a dead outbound link on an otherwise strong page gives the system a measurable reason to demote it for that query. This applies even to highly authoritative sites; Pew Research found 54% of Wikipedia pages contain at least one dead reference link.
How often should backlink health be monitored?
Given that roughly 7% of a backlink profile decays annually, a one-time audit only provides a temporary picture. Ongoing monitoring, checking referring domain health and AI citation status on a regular cadence rather than only during periodic cleanups, catches decay before it meaningfully affects visibility.
What's the easiest way to fix a dead backlink?
A dead inbound link is often caused by a moved rather than deleted page. Reaching out to the site owner with the correct current URL frequently restores the link with far less effort than building a new placement from scratch. This is typically the highest-return fix available, since the site already agreed the content was worth linking to once.
How do I find broken backlinks pointing to my site?
Most backlink tools (Ahrefs, Semrush, Moz) include a lost-links or broken-backlinks report that flags referring pages returning a 404 or redirect error. Sorting that report by the linking domain's authority helps prioritize which dead links are worth reclaiming first, since a lost link from a low-value site matters far less than one from a genuinely relevant, authoritative page.
Is link reclamation worth the effort compared to building new links?
Generally yes, for the links worth reclaiming. Reclamation targets a site that already decided your content deserved a link once, which removes the hardest part of link building (convincing someone your content is worth referencing) from the process entirely. It's typically faster and has a higher success rate than cold outreach, though it only applies to links that already existed and broke, not to earning placements on entirely new sites.
The Bottom Line
A backlink profile isn't a static asset once it's built. It decays at a measurable, predictable rate, and AI systems are actively built to notice and penalize that decay in ways traditional Google rankings mostly don't. The sites losing AI visibility right now often aren't doing anything wrong today, they're just running on content and links that were right two years ago and haven't been checked since. For the ranking-side version of this same erosion, see Why Google Is Ignoring Most of Your Backlinks.
Treating link health and content freshness as ongoing maintenance, rather than a one-time build, is what actually protects AI citation eligibility over time. If you want help auditing where your own backlink profile is quietly decaying, and fixing it before it costs you visibility, that's what we help clients with at LinkyJuice.
