Nearest Seed: How Google’s PageRank_NS Actually Works and Why It Doesn’t Replace PageRank

You've probably heard that PageRank is dead. It isn't. You may have also heard the newer version of the story, that a leaked attribute called PageRank_NS quietly replaced it. That one isn't true either.

Here's my stance up front: PageRank_NS is real, the math behind it is genuinely clever, and for most businesses it is not the number deciding your search rankings. Classic PageRank still is. But the seed distance model explains something the classic algorithm can't, which is how Google decides whose links to trust, and if you do link building for a living, or pay someone who does, that distinction should change which links you buy, chase, and ignore. This article explains it in plain English, at a level a ninth grader could follow.

The short version: Google was granted the nearest seed PageRank patent in April 2018. Three months later, the Medic update gutted the health niche. Google then spent years telling everyone the answer was E-E-A-T. Add author bios. Show your credentials. My read is that E-E-A-T was the decoy, the same way the Helpful Content Update was sold as a content quality system while the winners and losers sorted suspiciously well by brand strength. Classic Google, pointing the industry at the part it can see while the math runs where it can't. The measurable mechanism, the one with a patent number and a leaked PageRank attribute, is distance, your PageRank set by how far your pages sit from the nodes Google trusts most.

What PageRank_NS Actually Is

PageRank_NS is nearest seed PageRank, one of several PageRank variants named in Google's leaked documentation. The NearestSeeds method doesn't count the links pointing at your pages. It measures link distance: how far your pages sit from a small set of hand-picked, trusted seed sites in the web's link graph. Short distance from multiple seeds means high PageRank. And it doesn't replace the classic version. It sits alongside it, a trust layer on top of the same link graph.

NS stands for NearestSeeds. The attribute surfaced in the 2024 Google API leak, and its definition maps directly onto a Google patent, US9953049B1, "Producing a ranking for pages using distances in a web-link graph", filed by Google engineer Nissan Hajaj with a 2006 priority date and granted in April 2018. The leak gave us the attribute name. The patent gives us the ranking machine.

One honest caveat. The connection between the leaked pagerankNs attribute and this specific patent is a community inference, not something Google confirmed. The names match and the method matches, so I find it convincing. But it's an inference.

You'll also read that PageRank_NS is really about content quality, document clustering, or semantic relevance, an authority signal blended with content signals. Several popular leak breakdowns describe it that way, and I understand the temptation, because Google search does evaluate content relevance elsewhere. The patent doesn't support it. Nothing in the NearestSeeds math reads your content. The algorithm reads the link graph, measures link distance, and produces a ranking for the page. Quality content decides whether anyone links to your pages in the first place, which is upstream of this calculation, not inside it. Content relevance and content quality still matter enormously in Google search. They just live in other systems.

The Google Patent Behind the Nearest Seed Method

The original PageRank algorithm, the one Larry Page and Sergey Brin published in their 1998 "Anatomy of a Large-Scale Hypertextual Web Search Engine" paper, worked like an election. Every link was a vote. Pages with more links from more important pages earned more PageRank and ranked higher in search results, softened by a damping factor of 0.85 that modeled a surfer who eventually stops clicking.

The problem is that elections can be stuffed. Link farms, PBNs, and paid links all manufacture votes, which is why Google built the Penguin algorithm to catch link spam. Google's 2006 patent flipped PageRank instead of policing it. Stop counting links. Start measuring distance.

Think of a small town where everyone knows the sheriff is honest. Your reputation isn't how many people will say your name. It's how many handshakes you are from the sheriff. The patent picks a small set of "sheriffs," seed sites a human at Google manually reviewed and deemed reliable, high quality, diverse across topics, and heavily linked. The New York Times is the example the patent itself names. The old Google Directory is the other.

Roger Montti made a point at Search Engine Journal worth sharpening: this is not a "trust algorithm." Unlike the older TrustRank concept, which propagated an actual trust score from page to page, Google's patent never assigns trust to anything. Google picks trusted starting points, then measures pure distance. Trusted, in this usage, mostly means spam-free and high quality. The distance does the ranking, and the ranking is the PageRank. Distance can feed a trust decision without ever being a trust score itself, and that difference matters later.

Every page on the web then gets a PageRank based on its distance from those seeds, and the score decays exponentially. In the patent's math, your PageRank is proportional to e to the negative distance. Each extra hop away cuts the score by a big multiple.

The 2024 documentation shows PageRank_NS sitting alongside other PageRank variants inside Google search: rawPagerank, the plain link calculation. Pagerank2, an updated version the docs don't explain. FirstCoveragePageRank, the PageRank a page carries the day Google first finds it. And Toolbar PageRank, the public 0 to 10 toolbar number from the Google Toolbar, frozen in 2013 and retired in 2016. More than 14,000 attributes surfaced in that leak, and the same family of ranking signals kept showing up across them. Five variants in one leak is not what replacement looks like. It's what layering looks like: different PageRank flavors for different jobs, blended differently per query.

Why One Seed Is Never Enough

This is the part of the nearest seed method almost everyone gets wrong, and it's the single most useful thing in the patent.

Your PageRank in this system is not set by your distance to the closest seed. It's set by your kth-closest, where k is a small number greater than one. Not the nearest. The kth. The granted claims require it, and the patent's examples use single digits. So if k is 4, Google looks at the fourth-nearest trusted seed to your page and computes your PageRank from that link distance.

Why? The patent says it directly: using the kth seed suppresses "unfairly high scores due to lack of proportionality at the vicinity of the seeds." In ninth-grade English: standing next to one trustworthy site doesn't make you trustworthy. It might just mean you got lucky.

Run the logic. Say k is 4 and you land one incredible link from pages right next to a seed. Your first-nearest distance just shrank. Your fourth didn't move. Your PageRank barely changes.

Now say four separate trusted neighborhoods each sit a short path from your pages. Your fourth-nearest seed is close, and your PageRank jumps.

The analogy I use with clients: graduate committees don't admit you on one glowing reference letter, even from a Nobel winner. They want several independent references. One voice can be bought or fooled. Four are a pattern.

"Independent" is doing real work there. Twenty links that all trace back to the same seed's neighborhood fill one of your k slots, not twenty. You need short paths to genuinely distinct trusted clusters: a media cluster, an institutional cluster, an industry cluster. Different sheriffs.

How the PageRank Algorithm Measures Link Distance

Skip this section if you don't care about plumbing, but the plumbing is where the money is.

Link distance in this ranking algorithm isn't just hop count. Each link gets a length, and two things make links "longer": the base cost of every hop, and the number of outbound links on the source page. A page with a thousand outbound links is a weak conduit for PageRank. A page with five is a strong one.

Think water pressure: PageRank flows through the page graph like water through pipes, and a page with a thousand outbound links is a pipe drilled with a thousand holes. By the time the flow reaches your page, there's nothing left.

Notice where the penalty lands: on the source page, not on you. Linking out generously from your own site doesn't lengthen anyone's path into your pages, and it doesn't leak PageRank the way the old sculpting crowd feared.

The patent also explains why this ranking algorithm exists: computing full PageRank from many seeds was too expensive. Google's engineers noticed the authority flowing into a web page is almost always dominated by one strongest incoming link, so the computation keeps that term and throws the rest away. "Find my strongest endorsement" becomes "find the cheapest route from a seed to my pages," a shortest-path problem computers solve fast. That route is your PageRank.

The Medic Update Landed Three Months After the Patent

Google was granted this patent in April 2018. On August 1, 2018, it rolled out a broad core update the industry nicknamed Medic, because the damage concentrated so heavily in health, medical, and other YMYL categories that the lists of losing sites read like a hospital intake sheet.

The most famous casualty was Dr. Axe. draxe.com was one of the largest natural-health sites on the internet, holding classic PageRank most publishers only dream about, and it lost roughly half its search visibility almost overnight. Run that collapse through the seed distance lens and it stops being mysterious. Millions of links from wellness content, supplement affiliates, and recipe roundups put its web pages close to nothing Google trusts about medicine. Measured by volume, Dr. Axe was a giant. Measured by distance to medical seeds, the journals, institutions, and university sites a hand-picked list would actually contain, it was a long way from every sheriff.

Weeks before Medic, Google had also updated the Quality Rater Guidelines to lean harder on expertise, authoritativeness, and trust. I can't prove Google switched this exact patent on inside the algorithm for that exact ranking update. Nobody outside Google can. But the timing rhymes, the target profile rhymes, and the mechanism rhymes. Seed distance is precisely the kind of quality computation you'd layer into YMYL scoring, where whose PageRank vouches for you matters more than how many links do. My read is that seed distance plays into E-E-A-T far more than the industry gives it credit, and that it's weighted hardest where being wrong hurts people: health, money, law.

There's a pattern here worth naming. When the Helpful Content Update arrived in 2022, Google talked about people-first writing while the sites that recovered looked less like better writers and more like bigger brands. With Medic, Google talked about expertise while the mechanism that best fits the damage is PageRank math in the web's link graph. I don't think Google lied. I think it pointed everyone's attention at the slowest thing to fix, credentials and content polish, while the fast-moving variable was where your links put you. If you were rewriting author bios in 2019, you were working the decoy.

Why Most Businesses Should Worry About Real PageRank Instead

The part most explainers skip: for the average business, this probably isn't the ranking signal that decides anything.

The decay is the reason. A PageRank value falling off as e to the negative distance collapses fast. Most local service companies, small manufacturers, and regional B2B sites sit a long way from every seed, and so do the pages they compete against in search results. When a signal compresses toward zero across the entire ranking, it can't be the thing separating position one from position nine. Something else does that work, and the evidence says that something is mostly classic PageRank, still flowing link to link. You can see it in the wild. Pages with no seed proximity still rank on ordinary links. Link volume still correlates, and cheap links still sneak value through. Even spam links still produce lift before Google catches up with them, which no pure distance algorithm would allow. Traditional PageRank is visibly alive at the bottom of the market.

So the number to worry about is real PageRank, the kind that flows into your pages without getting suppressed or zeroed for making your site look like a link farm. What changed over the years is not the core idea but the plumbing that pushes authority through the web. The reasonable surfer model from Google's 2004 patent decides how much each individual link passes, weighting links by position, prominence, and topical relevance, a stand-in for how likely a real person is to click them. Links stopped being votes a long time ago. They're routes for authority, and some routes carry more of it than others. The context around a link now behaves like a relevance multiplier, with the leak showing Google stores the words on both sides of every anchor and demotes links that don't match their destination. And SpamBrain, Google's link spam algorithm, sits over the whole graph, neutralizing the links it decides are junk content or paid placements instead of penalizing you for them, quality control at graph scale.

None of this rescues the vanity metrics. DR, UR, DA, Authority Score: all third-party reconstructions of the original PageRank algorithm, built by counting links, and all inflatable by the exact tactics SpamBrain exists to catch. They estimate link quality from quantity, which is exactly the assumption Google abandoned. Buy enough high-DR placements from pages that sell them, and your DR climbs while your real PageRank doesn't move an inch. One is manufactured authority, the other is inherited authority. If you've ever bought a high quality guest post on paper, a placement on a DR 70 site with no real authority of its own, and watched your rankings do absolutely nothing, you already know these are different numbers. None of them is Google's PageRank. I wrote about that pattern in my fake SEO breakdown, and this math is the mechanism underneath it.

What the Nearest Seed Model Changes About Link Building

So who should do link building for distance? Anyone in a YMYL category, where the trust layer appears to carry real ranking weight, and anyone in an insulated niche where the web graph is sparse. Everyone else should treat it as link building hygiene that quietly compounds. Quality content earns the links. The web graph decides what that quality is worth. Either way, if the model is real, and I run link building as if it is, several standard practices flip.

Chase hubs, not scores. The links that move rankings most come from sites that are themselves already close to several distinct seeds: a major trade association page, a university department, a government spec reference, a serious industry publication. One link from a true hub hands your pages short routes to every seed near it, and several of your k slots fill from a single batch of links. A DR 70 blog with no real PageRank neighborhood fills none.

Diversify by cluster, not by count. Four links from four sites in the same PageRank neighborhood are one reference letter photocopied four times. Spread your links across genuinely different trusted regions of the web graph. Media, institutional, supplier, association.

Stop being afraid to link out. The out-degree penalty hits the source pages, never the destination. Linking to good pages costs you no PageRank in this ranking algorithm, and in an under-linked niche it can make your pages the connected node other sites route through. Hoarding PageRank is fighting the last war.

If you're in an insulated vertical, the math is on your side. Niche manufacturing sites, residential construction, industrial services: nobody links to anybody. Every seed is far from every competitor, which means ranking is decided by relative distance inside a sparse neighborhood. In a graph that empty, one legitimate association or supplier link moves your PageRank further than fifty links would move a SaaS company in a crowded market. I watched this on a manufacturing audit: hundreds of links from directories and scraper sites doing nothing, while the one link from a regional trade association did measurable work, because of where it sat in the web graph. It also has to sit on a crawlable, technically clean site, which is where technical SEO stops being optional. A short path Google can't crawl is not a path, and uncrawlable pages earn no PageRank at all.

Questions People Keep Asking

Does PageRank_NS replace regular PageRank? No. The leak shows multiple variants living side by side, and the aggressive decay makes kth-seed scoring most useful as a trust layer, likely weighted hardest on YMYL queries. The classic PageRank calculation still does the everyday ranking work.

Does Google still use PageRank? Yes. Gary Illyes confirmed the algorithm was still in use in 2017, and the 2024 leak listed multiple live variants, with pagerankNs described as the production value.

When did Google stop showing PageRank scores? The Google Toolbar stopped updating in December 2013 and was retired in March 2016. The public toolbar number died. The internal algorithm didn't.

Can you manipulate PageRank? You can manipulate the third-party imitations of PageRank, and people do, with link schemes Penguin was built to devalue. Manipulating seed distance requires convincing pages genuinely close to trusted sites to link to you, which is less manipulation and more just earning it.

Do you need links from the seed sites themselves? No. Distance is measured in short paths, not direct links. A stingy, well-connected page two hops out from a seed still puts your pages close.

What if nobody in my industry links at all? Then competing pages sit just as far from the seeds as yours do, and a single high quality link carries outsized weight in the ranking. Sparse graphs reward the one company that takes the work seriously.

The Distance Is a Layer

Fifteen years of link audits taught me one pattern: winning backlink profiles never look impressive in Ahrefs first. They look connected. A short, boring list of quality links from association sites, suppliers, institutions, and one or two real publications beats a thousand-row export of rented content placements, because Google isn't only counting PageRank votes anymore. It's also measuring PageRank as distance, checking how many handshakes your pages are from someone it trusts, and the closer your market sits to health, money, or the law, the heavier that measurement gets.

If you want your Nashville business measured by what Google actually computes instead of a vanity score, that's the work we do here.

Know which score your market runs on. Then build for it.