Transcript
Intro
Two links to the same site. One sits in a film critic's review, on a page with four links. The other sits on a directory page with two thousand. If both pages held the same rank, the original PageRank maths says the first passes five hundred times more. Same click, wildly different value.
And the link count is only the first multiplier. By the end of this episode you'll be able to look at any link and guess whether it carries real value, a little, or close to none, and then decide where that value goes once it lands on your site.
Authority as flowing water
Start in 1998, with the PageRank patent. A page's rank comes from the ranks of the pages linking to it, and each page splits what it passes across all its links. The patent's own example: page C gets a small constant, plus a quarter of A's rank, plus half of B's. The whole idea, in one line.
Think of it as water. Every page holds some and pours it out through its links. Four links, a quarter down each pipe. Two thousand links, a trickle down each. That's where our five hundred times came from: a quarter against one two thousandth, with both tanks equally full.
And water leaks at every step. The patent has a damping factor, because real people wander off after following a few links. So value fades with distance. A shiny link on a page that sits many hops from anything strong passes very little.
Picture a national newspaper linking to a small film blog, and that blog linking to ReelRadar. ReelRadar is now two hops from a very strong source, and some value leaked on the way. A direct link from the newspaper would be one hop. A link from a blog nobody links to is a hop from nowhere, which passes about what you'd expect.
Is any of this still running? The maths has surely changed since 1998, but the signal hasn't gone anywhere. A trial exhibit in the US antitrust case describes PageRank as a signal about distance from a known good source, feeding the quality score. Distance, again.
So where did PageRank is dead come from? In 2016 Google pulled the public PageRank score from its toolbar, and people mistook the fuel gauge for the engine. Google said it kept using PageRank internally, and the court record shows it still did years later.
Not every link pours equally
Back to the water. The simple model pours evenly into every link on a page. Real readers don't click evenly. A Google patent filed in 2004 describes a reasonable surfer, who follows some links far more often than others.
The patent lists what could set those odds. The anchor's font size. Where the link sits: in running text, above the first screen, in a footer or a sidebar. Its colour, including text the same colour as the background, which is exactly as sneaky as it sounds. And how many words the anchor has, and which ones.
You can check this before you pay for a link. Open the page where it would sit. Is it inside the article, near the top, in a sentence a reader cares about? Or is it in an author box, a partners list, or a sidebar widget that repeats on every page? In the patent's terms, those are the links people rarely follow.
A patent isn't proof anything is live. But the 2024 leak shows Google stores some of these features for each link: the anchor's font size, and a hash of the words near the anchor. The usual ceiling applies. The leak shows what Google stores, not how much it weighs or whether it's switched on.
Google's own link guide says it in plain words: the words before and after links matter. So the strong case is a link inside a sentence about the topic, in the main content. The weak case is a footer, a sidebar, or one more link in a long block of links.
Links are graded before they count
Now the part that should change how you buy links. In the leaked documents every link records a source type: the quality of the page it comes from. Three tiers. High, from what Google calls base documents. Medium, roughly the supplemental tier. Low, roughly what the documents call blackhole documents. Great name. Terrible place for your link.
The documentation even says a lower source type value means a more important anchor. Now picture a typical link package: hundreds of links from pages nobody visits, on sites built to sell links. Guess which tier those live in. You can't see the tier, but you can make a decent guess from whether the page gets real visitors.
A rough test you can run in ten seconds: search Google for the page's exact address with site colon in front. If nothing comes back, the page probably isn't indexed, and a link on it is decoration. Then ask whether the page ranks for anything and gets real readers. Those answers tell you more than any badge.
There's a second grade. When a link comes from a homepage, Google stores whether that homepage is not trusted, partially trusted or fully trusted. Put that next to the court record's distance from a known good source and the picture holds together. Trust starts somewhere and spreads through links, getting weaker as it goes.
One more stored field. For each source page the documents record outsites, the approximate number of sites it points to. A page that links out to hundreds of different sites is usually exactly what it looks like: a page built to hand out links.
The idea is older than the leak. In 2004 researchers from Stanford and Yahoo, not Google, published TrustRank. Hand check a small set of good sites, then follow links outward, betting that good sites rarely link to spam. Whether Google runs anything like it is unknown. What we do know is that Google cares about the same distance.
Which brings us to a popular claim: low quality links just get ignored, so cheap links can't hurt and might help. Half right. Google says SpamBrain neutralises unnatural links and the credit is lost. But the same spam policies say violating sites may rank lower or vanish. Ignored is the usual outcome, not a promise.
DR, DA and the scores you can see
Now the number everyone buys links by: domain rating, domain authority and their cousins. The claim is that these are Google's score, or close enough. They're not. They're estimates of the link graph, built by tool companies with their own crawlers.
The tool makers say so themselves. Moz calls Domain Authority a comparative metric and says it's not a direct ranking factor. The leak shows Google keeps its own site level authority value, and nothing in it ties to any third party score. Handy for comparing sites quickly. Folklore as a target.
And these scores are easy to inflate. A site can pile up thousands of junk links and show a lovely rating while every one of those links sits in the bottom tier. Sellers know buyers filter by the number, so the number is what they manufacture. Judge the page, not the badge.
The eight multipliers
So let's turn all that into something you can use. This is our working model of what a link is worth, built from the patents, the leak fields and our field work. It's not a Google formula. Eight factors, and they multiply rather than add.
Multiply is the important word. Add the scores, and a link that's great on seven factors and zero on one still looks decent. Multiply them, and one zero sinks the lot. A beautiful, relevant, in text link on a page Google never indexed is worth exactly nothing.
Run three common links through it. A guest post on a real, relevant blog with readers scores well on most factors. A post on a site that sells hundreds of placements a month fails independence and probably the tier. A general directory listing fails topic and placement. Most of what people buy fails at least one factor badly.
Now score a good one. A film festival's blog writes about a horror retrospective and links ReelRadar's horror streaming page in the second paragraph, with the page title as the anchor. Indexed, independent, on topic, in the text, natural anchor, a page with real readers. That's a link where every multiplier is pulling in the same direction.
Two factors that sellers never mention. Independence: a link from a site you own, or from a partner who links to you on every page, isn't a vote from someone else, and search engines have always tried to tell the difference. Topic match: anchors feed relevance in the court record, so a link from an unrelated site says little about what your page is about.
Factor eight is the one people learn about too late. In our field work, a good link landing on a profile that's mostly paid exact match links does less than the same link landing on a clean site. Links get judged by the company they keep. Cleaning that up gets two episodes of its own later in this part.
The scorecard also explains the advanced methods. Consolidating power from a strong, relevant old domain scores high on tier, topic and trust distance in one go. That's why it can work. And it explains why most attempts fail: the domain is off topic, its own links are junk, or the setup leaves a footprint that kills independence. Same scorecard, applied honestly. Precision work, not a shortcut.
And that's the maths behind one link outweighing five hundred. Value gets split by the number of links on the source page, graded by the source's quality, and faded by distance from trusted sites. Multiply those together, and five hundred weak links can add up to almost nothing.
Where the value goes next
Now the part most link builders ignore completely. Every external link lands on one URL. Where the value goes next is your call. It flows on through your internal links by the same rules: split across links, stronger in the main text, fading with every hop.
First, find out which of your pages actually earn links. Search Console's Links report has a table of your top linked pages. It's a sample, not the full list, but it's Google's own view. The answer usually surprises people: the homepage, one old guide, one data page.
Now route it. The pages that earn links are rarely the pages that earn money. Journalists link to ReelRadar's data page on how long films take to reach streaming. Nobody links to the where to watch pages that bring in the affiliate revenue. A couple of in text links from the data page fixes that.
Google's link guide sets the ground rules. Use real anchor elements with an href, or the link may not be followed at all. Make sure every page you care about is linked from at least one other page. And in Google's words, there's no magical ideal number of links. Sorry to anyone who was hoping.
Then think in hops. Value fades with distance, so money pages should sit one or two clicks from the pages that earn links, not five. On ReelRadar the where to watch pages were four clicks deep, behind year and genre archives. One contextual link from the data page made it one hop.
While you're there, look at your homepage. It usually holds the most link value on the site, so what it links to matters. ReelRadar's homepage gave prime space to the newsletter signup and the privacy policy, while the where to watch hub sat in a menu. Flip that around, and the hub gets the flow it deserves.
One old trick to retire. Some sites still nofollow their own internal links, to stop value leaking to pages like contact or login. That stopped working in 2009. The share on those links isn't handed to the others. It just evaporates. Link fewer things instead, and put the important links in the text.
And nofollow links from other sites? Since 2019 Google treats nofollow, sponsored and ugc as hints, not rules. It may still use them to understand links, including what the anchor words say about the target. So a nofollow link from a busy, relevant page isn't worthless. You just can't count on it.
Wrap up
The episode on one screen. A link's value is split by its neighbours, shaped by where it sits, graded by the page it comes from, and faded by distance from trust. Then it flows on through your own site, and that part is entirely yours to control.
The eight factor link value scorecard is free on the episode page at irongoo.com. Score your last ten links with it before you pay for the next one. Next up: anchor text, the lever most people either ignore or abuse.
