PageRank explained for modern SEO
PageRank is simultaneously the most famous and most misunderstood idea in SEO. The green toolbar number that made it household knowledge has been gone for years, which led many to conclude PageRank is dead. It isn't — the public score died, the algorithm didn't. The mechanism that propagates importance through a link graph is still the best model we have for how internal authority moves. Here's how it actually works and how to reason with it in 2026 without the cargo-cult baggage.
Run the Authority Flow Checker on your site — free, no account.
The toolbar is dead; the algorithm is not
Google retired the public PageRank toolbar score around 2016, and 'PageRank is dead' became received wisdom. The confusion is between a number and a method. What died was the visible 0–10 score Google chose to stop exposing. PageRank the algorithm — and the family of link-based importance computations descended from it — remains foundational to how a link graph is valued. You just can't read a page's score off a toolbar anymore.
Why it still matters to you: You don't need Google's exact numbers. You control your entire internal link graph, and PageRank tells you how authority propagates through it — which is exactly the lever you can pull. It's a model for your decisions, not a metric to chase.
How PageRank works
The core idea: a page is important if important pages link to it, recursively. The mechanic that makes that computable is the 'random surfer' — imagine someone clicking links at random forever. PageRank is the probability they land on any given page. Pages with many links from well-linked pages get visited more, so they score higher.
A links to B and C. B links to C. A ──► B │ │ ▼ ▼ C ◄───┘ C is the strongest: two pages link to it, and one of them (B) is itself fed by A. Importance is recursive — a link from a strong page is worth more than a link from a weak one. You can't value a link in isolation.
The damping factor
Real surfers don't click forever — sometimes they jump to a random page. PageRank models this with a damping factor (classically ~0.85): follow a link 85% of the time, teleport 15%. Damping stops authority circulating infinitely in loops or getting trapped in dead ends, and it's the reason a link's value decays with each hop it travels from its source — the mathematical root of why crawl depth and authority are the same problem.
What's changed since 1998
Modern ranking is hundreds of signals, and link-based importance is only one of them — so PageRank is necessary context, not a ranking shortcut. Two refinements matter for how you reason about it:
- The reasonable surfer. Google's patents describe weighting links by how likely a real user is to click them — a prominent in-body link counts for more than a buried footer one. This is the formal basis for valuing contextual links over navigation links, and it's already in how RankForge weights edges.
- Links are one input among many. Content quality, intent match, and trust signals all matter. PageRank explains how authority flows; it doesn't claim to explain ranking on its own.
- It's applied to a cleaned graph. Nofollow, canonicals, redirects, and duplicate URLs all reshape the graph before authority is computed — naive PageRank on raw links misreads real sites.
Why every page starts with something
One detail of the original model explains a lot of real-world behaviour: every page in the graph begins with a small baseline share, independent of what links to it.
In the model this comes from the random surfer occasionally jumping to a page at random rather than following a link. Practically it means a page with no inbound links is not at zero — it holds a small amount and passes it on. That is why an orphan page can still rank for something very specific, and why the fix for one is worth doing rather than hopeless.
- A page with no inbound links still holds a baseline and can rank for low-competition queries — it is starved, not dead.
- Adding pages to a site divides the baseline across more of them, which is one reason publishing thin pages at volume tends to weaken a site rather than strengthen it.
- Because the calculation runs to convergence across the whole graph, one new link changes every page's share slightly. Structure is a system, not a set of independent pages.
That third point is the practical reason a structural audit cannot be done page by page. Whether a link helps depends on what else exists, so the same link is a good idea on one site and a waste on another.
What PageRank is not
Most confusion comes from four different things sharing the name.
- Toolbar PageRank — the green 0-10 bar retired in 2016. It was a heavily rounded, months-stale snapshot even when it existed, and nothing has replaced it publicly.
- Domain Authority, Domain Rating and similar — third-party estimates built by Moz, Ahrefs and others from their own link indexes. Useful for comparing sites; not the algorithm, and not available to Google.
- The original 1998 paper — the published formula is the ancestor of what runs today, not a description of it. Two and a half decades of undisclosed refinement sit in between.
- Internal PageRank — the one you can actually compute and act on: the same maths run over your own site's link graph. Nobody has to publish anything for this to work, because you have the graph.
That last one is why the concept still matters despite the metric being gone. You cannot see Google's number, but you can compute how authority moves between your own pages, and that is the part you control.
Common misconceptions
“PageRank is dead.” The public toolbar score is gone; the algorithm and its descendants are alive and central to link-based ranking. Don't confuse the metric with the method.
“Every link passes the same PageRank.” Under the reasonable-surfer refinement, links are weighted by prominence and likelihood of being clicked. A contextual link and a footer link to the same page are not equal.
“More links to a page always raises its PageRank.” Authority a page passes is split across its outbound links, so adding links elsewhere dilutes. Net authority is conserved and redistributed, not minted.
What it looks like on a real site
The algorithm is easier to trust once you see what it implies about ordinary linking decisions. Three patterns show up on almost every site.
The homepage hoards
The homepage collects most external links, so it starts with the most to distribute. If it links mainly to the navigation and the navigation links mainly back, authority circulates around the top of the site and never reaches the pages that convert. It is the single most common finding in a structural audit.
The footer dilutes
A link's share depends on how many others sit beside it. A footer linking to sixty pages passes each of them a sixtieth of what it has to give, on every page of the site — a lot of links carrying almost nothing, which is why footer links rarely move anything.
Dead ends trap
A page that receives links and links nowhere accumulates authority and passes none on. One is harmless; a whole section of them — documentation, PDFs, terminal landing pages — is authority parked where it does no work.
All three are invisible in a page-by-page audit and obvious the moment you model the graph, which is what internal authority flow describes measuring.
Where the model stops matching reality
The random-surfer model is a simplification, and knowing where it diverges keeps you from over-applying it.
- Links are not equal. The original model treats every link on a page identically; modern systems weight by position, prominence and context, which is why a body link outperforms a footer link far more than the arithmetic alone suggests.
- Not every link is followed. Nofollow, and links behind interactions a crawler does not perform, remove edges from the graph entirely.
- Relevance is separate. PageRank says nothing about what a page is about. It decides which of several relevant candidates ranks, not whether a page is a candidate at all.
- Sites are not closed systems. External links arrive continuously and unevenly, so a snapshot of internal distribution is a moving target.
- Google runs far more than this. PageRank is one input among many, and no amount of internal linking substitutes for the others.
The failure mode to avoid is treating an internal authority score as a ranking prediction. It predicts which of your pages the site itself supports best — useful, and a different question from where a page will rank.
How to use it today
Treat PageRank as the mental model behind internal authority flow: link to important pages from your strongest, most relevant pages; keep them shallow; don't dilute with bloated link lists; fix the leaks that distort the graph. RankForge runs a weighted, damped PageRank over your cleaned internal graph to compute exactly this — see how RankForge calculates authority flow for the full model.
FAQ
Is PageRank still used by Google?
The public toolbar score was retired around 2016, but link-based importance computations in the PageRank family remain foundational to how Google values a link graph. The metric was removed; the method wasn't. It's one of many ranking signals, not the whole system.
What is the damping factor?
It models a surfer occasionally jumping to a random page instead of following a link (classically about 85% follow, 15% jump). It stops authority looping infinitely or getting trapped, and it's why a link's value decays with each hop it travels from its source.
Can I still optimise for PageRank?
Yes — internally. You control your whole internal link graph, so you can route authority by linking to priority pages from strong, relevant pages, keeping them shallow, and avoiding dilution and leaks. You optimise the flow, not a published number.
Does Google still use PageRank?
Google has confirmed that PageRank remains part of its core ranking systems, though it is one signal among many and the version running now is far removed from the 1998 paper. What retired in 2016 was the public toolbar display, not the algorithm.
Can I see my pages' PageRank?
Not Google's figure — it has never been exposed and the toolbar approximation is long gone. You can compute internal PageRank over your own link graph, which is arguably more useful: it shows how authority is distributed between your pages, which is the part you can change.
Is Domain Authority the same as PageRank?
No. Domain Authority is Moz's third-party estimate of a domain's ranking strength, built from Moz's own link index. PageRank is a Google algorithm operating at page level. They correlate loosely and are not interchangeable — and neither tells you how authority moves inside your site.
Does adding more pages dilute PageRank?
Adding pages divides the baseline share across more of them, so in a narrow sense yes. In practice it only matters when the pages are thin and add nothing — good pages attract links and earn their share back. The dilution argument is a reason not to publish filler, not a reason not to publish.
Do outbound links to other sites leak PageRank?
Some flows out, and it is rarely worth worrying about. Linking to genuinely useful external sources is normal editorial behaviour that search engines expect, and stripping external links to hoard authority costs more in credibility than it gains.
Is internal PageRank worth optimising if my site is small?
Below roughly fifty pages, not really. The navigation already reaches everything, nothing is more than two clicks deep, and there is not enough authority in the system for redistribution to change much. The constraint on a small site is almost always content and external links rather than structure.
Is PageRank still worth learning about if the metric is gone?
Yes, because the concept is what makes internal linking legible. Once you understand that authority splits across a page's outbound links and decays with distance, decisions like where to place a link and whether to add another to the footer stop being guesswork. The number was never the useful part.
Put this into practice
Run the Authority Flow Checker to see this on your own site, or run the full structural audit for the complete picture — both free, no account required.