PageRank: The Algorithm Behind Link Scoring
PageRank is the algorithm Google launched with in 1998: a method that assigns every page on the web a number expressing its importance – computed purely from link structure. This guide explains the mechanics: the original formula, the model behind it, how it is computed iteratively, and what of it still holds in 2026.
Scope: this article is about the algorithm and its history. How you steer link value in practice – internal distribution, redirect loss, sculpting – is covered in link juice and link equity.
1. The core idea
PageRank carries a principle from academia over to the web: a paper cited often is probably important – and citations from important papers count for more than citations from obscure ones. Larry Page and Sergey Brin applied that to hyperlinks.
At its heart sits a recursive definition: the importance of a page depends on the importance of the pages linking to it. That sounds circular, but it is solvable – section 5 shows how.
Two things made the approach revolutionary: it was query-independent, so a page’s value could be computed once in advance, and it was hard to manipulate, because you have to convince other people’s sites to link to you.
2. The original formula
In the original paper the formula reads:
PR(A) = (1 - d) + d × ( PR(T1)/C(T1) + ... + PR(Tn)/C(Tn) )| Symbol | Meaning |
|---|---|
PR(A) |
The PageRank of page A – the result |
T1 … Tn |
Every page linking to A |
C(T) |
The number of outbound links on page T |
d |
The damping factor, conventionally 0.85 |
The decisive part is PR(T)/C(T): a linking page does not pass on its full value but
divides it by the number of its outbound links. A link from a page with 5 links is
therefore worth far more than one from a page with 200 – at equal source strength.
3. The random surfer model
Behind the formula sits an intuitive model. Picture someone wandering the web at random: they start on an arbitrary page and keep clicking a random link. The PageRank of a page is then the probability that this surfer is currently on it.
Two properties follow immediately:
- Many paths mean a high value. The more routes lead to a page, the more often the surfer ends up there.
- PageRank is a probability distribution. All pages’ values sum to a fixed total. PageRank is therefore distributed, not created – nobody can bring more of it into the world by rearranging their own site.
4. The damping factor
The factor d = 0.85 models the surfer eventually stopping and typing a new address instead.
With 85% probability they follow a link, with 15% they jump elsewhere.
This is not just realism, it is mathematically necessary. Without damping the algorithm would face two problems:
| Problem | What happens | How d solves it |
|---|---|---|
| Dead ends | Pages with no outbound links swallow the value | The surfer jumps onward |
| Link traps | A group of pages links only to itself and hoards everything | The jump breaks the cycle |
The term (1 - d) is the floor: every page receives a base probability even with no inbound links
at all. That is exactly why no page can fall to zero.
5. How it is computed iteratively
The recursive definition is not solved algebraically but by repetition. Every page starts at the same value; the formula is then applied over and over until the numbers barely move – that is convergence.
An example with three pages: A links to B and C, B links to C, C links to A.
| Pass | PR(A) | PR(B) | PR(C) |
|---|---|---|---|
| Start | 1.000 | 1.000 | 1.000 |
| 1 | 1.000 | 0.575 | 1.275 |
| 2 | 1.234 | 0.675 | 1.291 |
| converged | 1.163 | 0.644 | 1.193 |
Two observations: C wins because two pages link to it. And B ends up below its starting value because A has to split its value across two targets. A web-scale graph needs a few dozen passes to converge.
6. Toolbar PageRank: rise and retirement
From 2000 to 2016 the Google Toolbar displayed a PageRank value from 0 to 10. That number shaped an entire generation of SEO – and was misleading even while it lived:
- It was logarithmic. Going from 5 to 6 demanded a multiple of what 4 to 5 required. A PR 7 was not “a bit better” than a PR 6 but an order of magnitude stronger.
- It was stale. Updates came only a few times a year, so the displayed value often lagged the internal one by months.
- It was coarse. Eleven steps for billions of pages – the internal number is a floating-point value with far finer resolution.
The visible score became a commodity: links were sold by toolbar PR. Google stopped updating it in 2013 and removed the display for good in 2016. The algorithm itself was not retired – only its public display.
What matters today: when a tool shows you a “Domain Authority” or “Domain Rating”, that is not PageRank. Those are reconstructions built on the vendor’s own crawl – see Domain Authority. Google has published no PageRank value since 2016.
7. From random to reasonable surfer
The original model assumes every link on a page is equally likely to be clicked. That is plainly untrue: a link in the body copy gets clicked far more often than one in the footer.
Google described the refinement in a patent as the reasonable surfer: links are weighted by their expected click probability. Factors that can play a role:
| Factor | Effect on weight |
|---|---|
| Position on the page | Body copy above footer or sidebar |
| Font size and prominence | More visible links carry more weight |
| Anchor text | Descriptive anchors suggest genuine relevance |
| Topical proximity | Does the target fit the source page context? |
This makes the simple division by link count from section 2 an approximation. The direction still holds – more outbound links mean less per link – but the split is no longer even.
8. What still holds today
| Claim | Status in 2026 |
|---|---|
| Google uses PageRank internally | Confirmed – as one signal among many |
| The value is publicly visible | False, not since 2016 |
| Links split value by count | Approximately; weighted by expected click probability |
| The original formula runs unchanged | False – heavily revised, details unpublished |
| PageRank alone decides rankings | False – relevance, quality and intent weigh more |
The practical consequence is unglamorous: links from strong, topically relevant pages help, and every extra outbound link on a page reduces the share of the others. Everything beyond that is speculation about a formula whose current form nobody outside Google knows.
9. Common misconceptions
- “PageRank is dead.” What was retired is the public display, not the algorithm. Google has repeatedly confirmed it still uses it internally.
- “Domain Authority is the new PageRank.” Third-party metrics are reconstructions from the vendor’s own crawl and feed into no Google assessment.
- “There is one PageRank per domain.” PageRank is defined per URL. The homepage typically holds a site’s highest value.
- “
nofollowredirects the value.” A nofollow link’s share is not redistributed to the others, it lapses – see link attributes. - “More pages mean more PageRank.” New pages only bring the floor value. Value comes from inbound external links, not from volume.
Pro tip: you cannot measure a page’s PageRank – but you can measure its
outbound link count, and that is the divisor in the formula. Rank-O-Saur lists every link on a page,
split into internal and external, with its rel attribute. Pages with unusually many
outbound links are exactly the ones whose individual links carry the least weight.
10. Frequently Asked Questions
Does Google still use PageRank at all?
Yes. What was retired in 2016 was only the public toolbar display. Google staff have repeatedly confirmed PageRank is still used internally – though as one of very many signals and in a heavily revised form.
What does the damping factor of 0.85 mean?
It models the probability that a surfer follows another link rather than stopping. Mathematically it prevents pages with no outbound links, or self-contained link groups, from hoarding the entire value.
Is Domain Authority the same as PageRank?
No. Domain Authority, Domain Rating and similar scores come from third parties and rest on their own crawls. They are approximations for comparison purposes and have no influence whatsoever on how Google assesses a page.
Does every URL have its own PageRank?
Yes, PageRank is defined per URL, not per domain. Within a site the homepage usually holds the highest value, because most external links land there; from there it spreads through internal linking.
Why can I not see my PageRank anywhere?
Because Google has not published it since 2016. Tools that display such a number are computing their own metric. There is no official way to query the actual internal value.
What is the difference between PageRank and link equity?
PageRank is the specific algorithm and its formula. “Link equity” or “link juice” is the practical umbrella term for everything a link passes on – including relevance and trust. The day-to-day application is covered under link equity.
Does adding lots of subpages generate more PageRank?
No. Each new page only receives the floor value from the (1 - d) term. Additional
value comes solely from inbound external links; internally you are only redistributing what is
already there.
Read next
