PBN Backlinks: Why Every Network Decays, and What to Do With One You Inherited
· Royking Niba
A private blog network is a group of websites controlled by one party for the purpose of pointing links at a site that party wants to rank. In practice the sites are almost always built on expired domains that still carry historical backlinks, dressed up with recovered or rewritten content so they read as independent publications. Google’s spam policies classify this as link spam, and name it directly: sites and links created primarily to manipulate rankings rather than to serve people. My position, having built these and having spent years since dismantling them, is that a network is not a strategy with a risk attached. It is a depreciating asset with an unwind cost that grows every month you leave it running.
I can describe the machinery because I operated it. At Blue Window Ltd I ran the full private network lifecycle: expired-domain acquisition, archive.org content recovery, site restoration, and ongoing network quality monitoring. That was the job, and I was good at it. Most writing on this subject comes from one of two places, someone selling links or someone who has only read about them, and both produce the same vague article. Since then, PBNs that stopped working have become one of my named recovery specialisms, which means I now spend my time on the other side of the same problem: sitting with an owner who has inherited a site, discovered a network under it, and needs to know whether the thing holding up their rankings is also the thing about to take them away. Nothing below is a build guide. There are no operational instructions here, and deliberately so.
The lifecycle, described from the inside
Every stage of network operation exists to solve one problem: making a link look like it was earned. Understanding what each stage is really doing tells you why the whole structure eventually fails.
Expired domains are bought for their history, not their name
A brand new domain is worth nothing to a network, because the value being purchased is not the address. It is the accumulated link graph pointing at that address, built by other people over years for reasons that had nothing to do with the buyer. A university department that closed, a local newspaper that folded, a hobby forum that went dark. The links to those sites survive the sites themselves for a while, sitting on pages nobody has updated. Acquisition is the act of buying that residue.
This matters more than anything else in the article, so I want it plain: the authority in a network is borrowed from dead publishers, and it was never renewed. Nobody links to a restored corpse. From the moment of purchase the input signal only shrinks.
Archive.org recovery is continuity work
Content recovery pulls the domain’s former pages back out of the Wayback Machine so the restored site resembles what those old links originally pointed at. The purpose is continuity: the URLs that were linked should resolve to something plausibly like what was linked. It is not content production in any meaningful sense, and no reader was ever the audience for it. I wrote Python to do this at volume, which tells you what kind of task it is.
Restoration is set dressing
Restoration means reassembling a functioning website on that domain: a CMS, a theme, a menu, a plausible posting history. The site has to survive a human glance. It does not have to survive a reader, because it will never have one. Compare that to creating helpful content and the gap is not subtle. These sites fail the first question that guidance asks, which is whether the content exists to help somebody.
Quality monitoring is monitoring for detection
This is the stage that gives the game away, and I say that as the person who used to run it. Network quality monitoring watches whether pages are still indexed, whether authority metrics are sliding, whether the money site still responds when a link is added, and whether the network’s patterns have drifted into something recognisable. Every item on that list is a detection signal. Nobody running an actual publication monitors their portfolio for whether it has been caught yet. The dashboard is an admission of what the assets are.
Why networks work, and then stop
Networks decay for three reasons at once, and the reasons compound. Two of them have nothing to do with Google, which is the part most operators miss while they wait for an update to blame.
The borrowed authority rots. The historical links that made an expired domain worth buying are on pages that are themselves ageing. Those pages get rewritten, pruned, redirected into oblivion, or hosted on sites that shut down. Nobody is adding new links to replace them. A network’s raw input therefore declines from day one, silently, on a curve nobody plots because no operator instruments it.
The footprint converges. Networks get more uniform over time, not less, because uniformity is how operators keep costs down. The same host, the same theme, the same plugin set, the same publishing rhythm, the same anchor conventions, the same outbound pattern. Every efficiency added makes the network fractionally more legible as a set. This is the failure nobody designs in and everybody arrives at.
Google’s side changed from punishment to neutralisation. This shift is widely misread. In the earlier era, link spam that was caught produced a message in the manual actions report. Loud, dated, unambiguous. The far more common outcome now is that the links simply stop counting. No message, no date, no report. Rankings drift down over months and the operator concludes the market got more competitive. A fully devalued network and a working one produce identical dashboards, right up until you look at the rankings.
| Phase | What the operator sees | What is actually happening |
|---|---|---|
| Early | Links added, rankings move within weeks, causation feels obvious | The borrowed equity is at its maximum and has not yet been discounted |
| Plateau | New links move rankings less. Blamed on competition or on needing “stronger” domains | Diminishing returns as the network’s pattern becomes more consistent and less credible |
| Drift | Slow decline with no event to point at. No manual action. No update announcement that fits | Neutralisation. The links are still visible in every tool and are carrying nothing |
| Collapse | A sharp drop, often coinciding with a spam update, blamed entirely on that update | The update removed the residual, but the asset had been depreciating for a year or more |
The reason this pattern is worth naming is that it changes the diagnosis. If you find a network under a site that dropped, the network is rarely the sole cause of the drop on the day it happened. It is the reason there was nothing underneath when the support was removed. That distinction drives everything in Google penalty recovery, because a site with genuine demand recovers on a different curve to a site that never had any.
The tell: spotting a decayed network in an inherited profile
If you have taken over a site and suspect there is a network under it, you are usually looking without documentation. The previous owner or agency rarely hands over a list, and often genuinely does not have one. So you read it out of the backlink profile, where a decayed network has a very particular shape.
The strongest single tell is a content age that does not match the link age. A restored domain has a site whose oldest posts date from the year it was rebuilt, but a backlink profile whose links date from five or ten years earlier. Real sites accumulate links after they publish. Restored sites accumulate content after they were linked. That inversion is very hard to fake at scale and it is the first thing I check.
| Signature | How to confirm it | Strength |
|---|---|---|
| Linking site’s content is newer than its own inbound links | Compare oldest post date against the date distribution of that domain’s referring pages | Highest |
| Referring domains acquired in a tight window, then flat | Referring domain growth chart. A block, then a cliff, then nothing | High |
| Linking domains have authority metrics but no organic traffic | Any rank tracker. High DR, near-zero ranking keywords, is not a natural combination | High |
| The same linking sites also link to unrelated commercial sites | Check the outbound links of three or four suspect domains. Look for unrelated verticals | High |
| Shared hosting, theme, plugin set or WHOIS pattern | Open five of them side by side. Convergence is visible without tooling | High |
| Thin archives: twenty to sixty posts, then silence | Browse the archive pages. Publishing stops when the domain stopped being fed | Medium |
| Anchor text tidier than any real publisher writes | Anchor distribution report. Commercial phrases where brand and bare URLs should dominate | Medium, decisive in combination |
| Historical links to the linking domain now 404 | Spot-check the referring pages. Dead sources mean the borrowed equity has already rotted | Medium, tells you how far gone it is |
One of those signals is weather. Four of them on the same set of domains is a system. The sampling method I use to get through a large profile without reviewing every link is the same one in toxic backlinks, and the important discipline is identical: you are grouping by shared origin, not scoring links one at a time.
When the network is the reason you rank
Here is the situation that actually brings people to me. The audit is done, the network is confirmed, and the site’s commercial rankings are visibly resting on it. The instinct is to rip it out. That instinct kills sites: you remove the links and the rankings in the same week, and you have no traffic left to fund the rebuild you now need.
Rebuild first, then retire. Not the other way round, and not both at once. The sequence I use:
- Check for a manual action before anything else. If there is one, the sequencing choice is gone. You have to clean up genuinely and file a reconsideration request, and a partial cleanup will fail it. Everything below applies only when there is no manual action and the loss is algorithmic.
- Measure how much of your ranking is actually network-fed. Segment by page and by query. The dependency is usually concentrated: a handful of commercial pages carry it while the informational side stands on its own. That changes the size of the problem, usually downward.
- Assume some of it is already dead. If the profile shows the drift pattern above, a share of those links stopped counting a while ago. You may be about to remove links that are contributing nothing, in which case the risk you are agonising over does not exist.
- Build the replacement while the network still stands. Real coverage on the pages that matter, internal linking that pushes authority from your genuine pages to your commercial ones, and links earned by being worth citing. This is the slow part and it is why you do it first.
- Retire in tranches, smallest contribution first. Four to six weeks between tranches, with a fixed cohort of tracked queries measured across each gap. Move slowly enough that cause and effect stay readable. If a tranche produces a drop you cannot absorb, you stop and rebuild more before continuing.
- Never let the network die by accident. Lapsed domain registrations and expired hosting produce exactly the simultaneous collapse you are working to avoid. If you are running one down, run it down deliberately.
On timing, be realistic. When I recovered roughly 45,000 monthly organic visits for a sports-media property after a spam update, the work ran across an August to October window. A quarter is a normal shape for this kind of recovery. Anyone promising you a fortnight is selling something.
Should you disavow your own network?
Usually not, and the reason is that disavowing is the weaker of the two tools available to you. If you control the linking domains, you can remove the link, noindex the page, or retire the site outright. That is a real change to the web. The disavow links tool only asks Google to ignore something, and Google’s own guidance is that most sites never need it. Filing a disavow for a network you still own and still operate is theatre, and it does nothing that switching the links off would not do better.
There are two situations where I do file. The first is a manual action for unnatural links, where the reconsideration request needs to show demonstrable effort and there are domains in the set you genuinely cannot reach: sold on, lapsed, controlled by a former agency who has stopped replying. Disavow is the honest answer for those, alongside removal of everything you can actually remove. The second is a network you inherited and do not control at all, where the previous owner kept the domains. Then disavow is the only lever you have.
What I would not do is pre-emptively disavow a network in the absence of a manual action, purely to feel clean. If the links have already been neutralised, the file changes nothing you can measure and you have removed your own ability to tell what was carrying weight. The inclusion test, and the cost of over-disavowing, are set out in full in the disavow file explained.
| Your situation | What I would do |
|---|---|
| You control the domains, no manual action | Retire in tranches. Do not disavow. Removal beats disavowal |
| You control the domains, manual action issued | Remove genuinely, disavow the remainder, then file the reconsideration request |
| You inherited the site, someone else holds the domains | Disavow, because it is the only lever you hold, and rebuild in parallel |
| Network confirmed but rankings do not depend on it | Cut it now. The optionality is worth nothing and the liability is real |
| Suspected network, evidence is one or two weak signals | Do nothing yet. Gather more evidence. Wrong amputations are permanent |
What is left when it goes
Whatever the site would have ranked for anyway. Owners find that hardest to hear before the fact and easiest to accept afterwards, because the residual is usually larger than they feared. Sites with a network under them are rarely nothing but a network. They have products, pages, some real coverage, and a brand people occasionally search for by name. The network was borrowing a position the site had not yet earned, and charging interest in the form of an unwind you would eventually have to pay.
The uncomfortable part is that the interest compounds. A network bought five years ago is harder to leave than one bought last year, because five more years of content strategy, internal linking and audience building did not happen. That opportunity cost is the real price of a PBN, and it never appears on the invoice.
The short version
- A PBN is a controlled set of sites, usually restored expired domains, that exists to pass links. Google’s spam policies classify this as link spam.
- The authority is borrowed from dead publishers and is never renewed, so it declines from the day of purchase whether or not Google ever notices.
- The modern failure mode is silent neutralisation, not a manual action. No message arrives. The links stay visible and stop working.
- The clearest tell in an inherited profile is a linking site whose content is newer than its own inbound links.
- Rebuild before you retire, retire in tranches with weeks between them, and never let the network lapse by accident.
- Removal beats disavowal when you control the domains. Disavow when you cannot reach them, or when a manual action forces the issue.
I built these for a living, and now I unwind them for a living. The second job is longer, more expensive, and far less fun than the first. That asymmetry is the whole argument.