Somewhere in a post you published, a citation now opens onto nothing. A 404, a login screen, a spinner that never resolves, a domain that answers to nobody. The sentence it was supporting reads exactly the way it did the day you shipped it: a number, stated as fact, with your name at the top of the page.
Swapping in a fresh URL takes five minutes. The expensive part is everything around it: knowing which of your claims just lost its evidence, which other links on the same page are dead too, and how long all of this sat in public before you happened to click.
Link Rot Reaches Your Citations
Link rot in a citation means a source you cited stopped answering while your claim stayed published. The URL starts returning a 404, a login wall goes up, or the domain lapses, and the sentence that leaned on that source keeps asserting its number to every reader who arrives. The link is the visible part of the failure. The claim it was holding up fails at the same moment and shows nothing.
Most advice files this under hygiene: sweep the page, patch the red rows, done. That handles the link. The harder problem it leaves standing is that you probably keep no record of which sources support which sentences, so when a source dies you can't say what it took down with it.
What the Status Codes Mean
Run a broken-link sweep and you get a column of red. A 403 sits next to a 404 sits next to a timeout, all wearing the same label. Each one is asking you for something different, and three of the four need no new source at all.
The 403 is where sweeps do the most damage. The page is alive, and a person with a browser can open it; your checker got bounced by a bot-wall. Treat that row like a 404 and you throw away a citation that was fine. Open it by hand first, and if it really does gate, re-link to a version that doesn't, or quote the figure and point at an archived copy.
Check Every Link on the Page
You want the whole picture in one pass, sorted by what actually failed. LiquiChart's broken-link checker takes the URL of a published post, pulls every outbound link, and reports each one as dead, gated, redirected, or timing out. It's a snapshot of the page at the moment you run it.
The dead rows are the ones you came for. Hold onto the green rows too: a 200 means the page loaded, and that is all it means.
Most dead citations get found by someone other than the author, months after the page went dark. Be honest about your own setup: when a source you cited dies, what actually tells you?
Every answer on that list is really a fuse length: how long a source can stay dead before the news reaches you. The first three are all versions of waiting, each with a slower burn than the last, and the waiting is the whole exposure. Whether a reader finds the 404 before you do comes down to that one number.
The Real Dead Citation Rate
How big this problem looks depends on who's counting. In our own corpus, a raw checker flagged 28.3% of linked citations as dead, and a real-browser recheck of the same links brought it down to 20.4%. The difference is bot-walls and consent redirects the raw checker couldn't get past, every one of them scored as a dead page.
So the honest figure is that about 1 in 5 linked citations is dead, gated, or broken. Inside that 20.4%, the genuinely dead pages, the 404s and 410s, come to 2.0%. Everything else is gated, redirected, or rate-limited, pages a person can still open even though a script bounces off, and any tool that only knows how to spot a 404 reads that bounce as a death. The same pattern runs through the state of content decay: publishers cite more than they used to, and verify about as rarely as they ever did.
I've run a sweep that came back all green and felt better than I should have. A status check clears the loud failures, and about the link that loads fine but no longer says what I cited it for, it has nothing to offer.
When the Page Loads but the Claim Is Gone
The riskiest link on your page returns a 200. It renders, the design looks current, and the paragraph you cited got rewritten in an update eight months ago. A status sweep waves it through every time, because the only question a sweep knows how to ask is whether the page answers.
The Content Health Scanner opens the live source and checks whether it still states the claim you cited it for. What that layer catches is its own subject: a link that resolves and still does not back your claim.
Put Cited Sources Under Watch
A sweep, a hand-check, a scan: each is a check you run once, and it starts aging the moment it finishes. Sources keep dying on their own schedule, whether or not you rechecked in March. I found a dead link in one of my own top-traffic posts during an audit, with no way to tell whether it had been dead for a week or for a year.
Whichever way readers above answer, every method shares one weakness: it runs on someone happening to look. The length of that delay is how long an unsupported claim sits in public with your name on it, and a schedule shortens it by running on a clock rather than on someone noticing. The fuse stays lit either way. A standing watch only decides how much of it burns before you reach the claim.
A watch won't keep your sources alive. It decides who finds out first, and how much daylight passes before you do. Add a cited source as a Monitored Page and LiquiChart re-checks it on a schedule, the same way it watches a live page for a changed number.
When a watched source comes back 404, 403, or 401, the claim that cited it gets flagged for review, and a Pulse beat lands on your timeline naming the page and the status it returned. A 5xx or a timeout raises a beat telling you to recheck, without flagging the claim, since a server that errors today often answers tomorrow. Nothing gets rewritten and nothing publishes on its own; the flag hands you the exact sentence that just lost its support, and the decision stays yours.
Who Finds Out First
A 404 shows itself to whoever clicks. Until someone does, a dead source keeps standing under a published sentence with your name over it, and that someone is usually a reader, a competitor, or a crawler. Silently unsupported claims are one of the ways blog traffic drops with no algorithm update anywhere in the story.
Some of your sources are dead right now. What's still open is the delivery: whether the news arrives as a line on your own timeline, or as an email from a stranger pointing at a number you can no longer stand behind.