PBN backlinks don't fail one at a time

Google's spam policy documentation does not contain the word "PBN". It does not contain "private blog network" either. I checked on 4 September 2026, on a page Google had last updated on 28 August. Nearly every page ranking for the term writes as though the phrase came from Google.

That is more than a pedantic correction. The category you are worried about is not a category Google names, which means no report will ever tell you "this link came from a network". There is no notification, no flag in Search Console, no line item. What exists instead is a shape in your own data: several links, on sites that look nothing like each other, changing state inside the same short window.

What is a PBN backlink?

A PBN backlink is a link from a site inside a private blog network: a group of websites controlled by one operator and maintained to sell links that pass ranking credit. The sites are usually built on expired domains carrying leftover authority from a previous life, and they are presented to buyers as independent publishers. What that leftover authority is actually worth by the time it changes hands is its own question.

The 2026 complication is that this classic shape is now the smaller half of the problem. The larger half is the general-purpose placement site: indexed, real traffic, competent content, a public "write for us" page, and a price list that sells to anyone with a card. It is not hiding a network. It is an open link shop. Both sell the same product and both fail in similar ways, but only one is trying to look like something else, and only one can be caught by looking for concealment.

Do PBN links still work?

Sometimes, for a while, and you will usually not be able to prove which.

Google's documented mechanism is not punishment. It is subtraction. Announcing the December 2022 link spam update on 14 December 2022, Google wrote that it was "leveraging the power of SpamBrain to neutralize the impact of unnatural links on search results", and that "ranking may change as spammy links are neutralized and any credit passed by these unnatural links are lost". The same post says the system detects "both sites buying links, and sites used for the purpose of passing outgoing links".

Read that closely, because it describes an outcome with no symptoms. If the credit a link passed is removed, your rankings do not fall below where they started. They stop being propped up. There is no message, no drop you can attribute, and nothing separating "this link was neutralised" from "this link never did much in the first place". That is the honest answer, and it is why the argument never ends: both sides are reading data that cannot settle it.

The footprints that give a network away

Most of the tells still being published stopped working. The single most important one to retire is content quality. Network content is written by the same models everyone else uses and reads perfectly well, so "it reads badly" is no longer a discriminator and any method resting on it is dead.

What still works looks at behaviour across a site rather than the quality of one page. Three consecutive posts linking to a dental clinic, a crypto exchange and a casino answer the question by themselves; sites that publish because they have something to say link out rarely, and to things they care about. So does one commercial link per post, in every post, without exception. So does cadence: nine posts in 2019, silence, then sixty last quarter, or output that begins the month the domain changed hands. A category structure spanning home, finance, health, travel and casino means the site takes anyone.

The strongest single check is the site's own referring domains. If its authority comes from thirty domains that also link to each other, you have found the network rather than a member of it. LinkGuard cannot do this one: it needs a link index, and ours is not one. That check belongs to the crawler suites, and on this particular question they are the right tool.

One question this article will not answer is which specific sites are networks. Naming a domain on the strength of a pattern is an accusation rather than a finding, and this pattern is right often enough to be dangerous. Describe the shape, then check your own list. The operational version of that screening lives in our donor evaluation checklist, and reading the seller rather than the site is covered in how to vet a link seller.

The pre-purchase version fits in one email. Before money moves, ask which URL the link will sit on, ask to see three recent posts from that same section, ask whether the placement is permanent or renewed, and ask what happens if the post is removed. The answers, and how quickly they arrive, tell you more than any score will.

Are PBN backlinks safe?

Google's spam policies, read on 4 September 2026, define link spam as "the practice of creating links to or from a site primarily for the purpose of manipulating search rankings", and list among the examples "exchanging money for links, or posts that contain links" and "advertorials or native advertising where payment is received for articles that include links that pass ranking credit". The stated consequence is that "sites that violate our policies may rank lower in results or not appear in results at all", detected "both through automated systems and, as needed, human review that can result in a manual action".

Buying a followed link therefore sits squarely inside the documented definition, with two documented outcomes. Neutralisation is the ordinary one, and by the mechanism described above it is silent. A manual action is the rare one, and it does produce a signal, in Search Console, under its own name. Which applied to your link, or whether either did, is not observable from outside. The wider version of this argument is in should you buy backlinks, and what a scoring tool means when it calls a link toxic is in what is a toxic backlink.

A PBN doesn't fail one link at a time

A guest post on a real site fails alone. The publisher redesigns, an editor honours a takedown request, a contract lapses: one link, one cause, one date. That is the story of most link loss, and we told it generally in how long paid backlinks last.

A network link was never really a link on a site. It was a link on one operator's asset, and that operator makes one decision at a time for everything they own. When a batch of registrations is left to expire, when hosting is migrated, when the network is sold, or when a footprint spanning every property finally catches up with all of them, the effect lands across the whole portfolio inside roughly the same window. Including on the four domains in your spreadsheet that look, to you, completely unrelated.

The failure also has an order to it. The page usually stops being indexed while still serving a clean 200: the URL loads, your link is still in the HTML, a checker still reports it found, and nothing is passing through it. The 404, the lapsed registration, the redirect into somewhere you would rather not be associated with, those come later. Sometimes much later.

Which turns the useful signal inside out. It is not "one of my links died". It is several links, on donors with nothing visible in common, changing state in the same week. Seeing that requires no hosting fingerprinting and no index. It requires your own dates.

Be clear about what this claim is. It is reasoning from how these networks are assembled and operated, not a measurement. We have published no decay data, our first study is not due until Q4 2026, and until it exists this is a shape to watch for rather than a rate to quote. Innocent explanations exist too: an index refresh, a monitoring failure on our side, two donors that share a host without sharing an owner. A cluster is a reason to go and look. It is not a verdict.

What that looks like in a monitoring log

On a link LinkGuard is watching, the sequence usually reads like this. The index check flips from indexed to not indexed while the status stays active. Then, weeks or months later, the link goes missing on a page still returning 200, or the donor stops responding at all, or a redirect appears and the chain ends somewhere unrelated to the site you paid. Our backlink index checker and redirect checker run those two checks on a single URL without an account, and the indexation article covers what a flip does and doesn't prove.

Those states are worth keeping apart. A link recorded as missing means the page was fetched and the link was not in it. A donor that will not respond is a different fact with a different cause. Neither is the same as a page quietly dropping out of Google's index while everything else stays intact.

Here is what we cannot see, stated plainly, because the alternative is a monitoring vendor overselling itself. A post that has been quietly orphaned, dropped off the category pages, no longer linked from anywhere on its own site, still returning 200 with your link and anchor intact, is invisible to a status check. LinkGuard scores that link healthy, because by every signal a tool reads from outside, it is. The only hint is indirect and late: buried pages tend to fall out of the index eventually, so an index check flipping on a page that is still live and still carrying your link is usually this, rather than anything done to you on purpose.

Grouping is manual today. LinkGuard alerts on each link on its own and does not tell you that four of them changed together. Export your links, sort by donor domain and by the date each one last changed state, and read down the column. That is the entire method, and a spreadsheet performs it.

If you already bought them

Usually, do nothing. That is the answer most pages ranking above this one will not give, because their recommended action is to buy safer links from the company that wrote the page. We sell monitoring rather than placements, which removes that particular incentive and leaves us with a different one worth naming: we benefit when you keep watching links you already paid for.

If the credit was neutralised, it is already gone, and removing the link changes nothing. Disavowing is a real instrument for a narrow case, mostly an unnatural-links manual action or a genuine negative-SEO campaign, and it is not a hygiene routine; our disavow checklist covers when it actually applies. If a donor now redirects somewhere unpleasant, trace the chain and judge that one link on its own merits. If the page is still live and the link still followed, you are holding an asset of unknown value, which is not an emergency.

One question here gets declined rather than answered: how many sites a network needs. That is a question about building one.

Questions people ask

What does PBN stand for?

Private blog network. The three letters are reused by several unrelated industries, so searches for the bare acronym return a mix of subjects; the spelled-out phrase is the more reliable search term.

How do you check if a site is a PBN?

Look at behaviour across the site rather than the quality of one page: who it links out to across consecutive posts, whether every post carries exactly one commercial link, its publication cadence, whether its own referring domains link to each other, and what the Wayback Machine says it used to be. No single signal is conclusive, and no tool returns a verdict.

Does Google penalise PBN links?

Google's spam policies never use the term. They describe link spam, and two outcomes: automated neutralisation, which removes the credit a link passed and produces no visible signal, and a manual action after human review, which appears in Search Console. Which one applied to a given link is not observable from outside.

What is the difference between a PBN and a guest-post site?

Intent and concealment. A network exists to sell links and is built to look like independent publishing. A site that accepts paid guest posts may be an ordinary publisher with a commercial sideline. In 2026 much of the paid inventory sits between the two: real sites, real traffic, a public price list, and no pretence.

Should I disavow a link from a PBN?

Usually no. Disavowal addresses a manual action or a deliberate negative-SEO campaign, not links whose credit an algorithm has already discounted. Removing a neutralised link recovers nothing.

The pattern is the finding

Nothing in a monitoring log says "network". The word does not appear in Google's documentation, in our status codes, or in any report you will ever be sent. What appears is a set of dates.

A single dead link tells you a publisher made a decision. Four dead links across four unrelated donors, inside the same fortnight, tell you one person did, about property they all belong to. The first is ordinary attrition. The second is the only view of a network anyone outside it actually gets.