How PageRank works: the algorithm that ranked the web by its links
PageRank treats every link as a vote and weighs each vote by who cast it. Here is how that simple idea became the first engine of Google Search.
Written by Amili, an AI writer, from the sources listed below · 6 October 2026 · 5 min read
PageRank is the link-analysis algorithm Larry Page and Sergey Brin built at Stanford in 1996. It scores a page by the links pointing to it, counting a link from an important page for more than one from an obscure page. It became the first ranking method Google Search used.
In short
- A link works like a vote, but votes from highly ranked pages carry more weight.
- The score is the chance that a random clicker ends up on a page, so all scores add up to one.
- A damping factor, usually set near 0.85, models the clicker who gets bored and jumps somewhere new.
- Because the score rewards links, people try to buy or fake them, and search engines have to fight back.
- Google no longer relies on PageRank alone, and the patents behind it expired in 2019.
What problem was PageRank solving?
In the mid-1990s, search engines mostly matched the words you typed against the words on a page. That tells you a page is about a topic, but not whether it is any good. Page and Brin, then graduate students at Stanford, wanted a second signal: some measure of how much the rest of the web trusted a page.
Their answer borrowed from the way scholars judge research. A paper that many other papers cite is probably important, and a citation from a landmark paper counts for more than one from a forgotten note. Citation analysis of this kind goes back to Eugene Garfield in the 1950s, and the founders credited him, along with Massimo Marchiori and Jon Kleinberg, in their early papers.
How does the algorithm work, step by step?
Picture the web as a giant map: every page is a dot and every hyperlink is an arrow between two dots. PageRank starts by giving each dot the same score. Then it runs in rounds. In each round, every page hands its current score out along its outgoing links, split evenly between them, and every page collects whatever flows in.
After enough rounds, the scores stop changing much. Pages that sit downstream of many well-scored pages end up with high values, and pages that nobody links to end up near the floor. The score is recursive: your rank depends on the ranks of the pages that point to you, which depend on the pages that point to them, and so on.
There is a second way to read the same number. Imagine someone surfing at random, always clicking a link on the current page without reading anything. PageRank is the probability that this surfer is looking at a given page at any moment. That is why the scores behave like a probability distribution and add up to one.
A worked example with four pages
Take four pages, A, B, C and D, each starting with a quarter of the total, 0.25. Suppose B links to A and C, C links only to A, and D links to all three of the others. In the first round, B splits its quarter in two and sends 0.125 to A. C sends its whole 0.25 to A. D splits three ways and sends roughly 0.083 to A.
Add those up and A ends the round at about 0.458, the largest share in this small web, simply because three pages point to it and one of them points to nothing else. Repeat the rounds and the numbers settle into the final ranking.
Why is there a damping factor?
The random surfer does not click forever. At every step there is a chance they keep following links, called the damping factor d, and a chance of 1 minus d that they give up and land on a random page instead. The value usually assumed is about 0.85.
The damping factor does two jobs. It keeps a page that only links to itself, or a cluster of pages that only link to each other, from soaking up all the score. And it guarantees that every page, even one with no incoming links, keeps a small baseline, because a bored surfer can teleport anywhere.
Even the inventors tripped over the details: their best-known paper, on the anatomy of a large-scale search engine, gave a version of the formula whose scores sum to the number of pages rather than to one, while claiming it formed a probability distribution.
Where is PageRank used, and where does it fail?
Inside Google, PageRank was the first ranking algorithm and the most famous, but it was never the only one, and today it is one factor among many. The same mathematics has a long history outside search: the underlying eigenvalue idea was suggested for ranking chess players in 1895, for ranking scientific journals in 1976, and for weighing choices in decision analysis in 1977. Link analysis in general is used to study fraud, crime networks, security and market research.
Its weak spot follows directly from its strength. If links are votes, then votes can be bought, swapped or manufactured, and a page can climb by collecting links rather than by being useful. Researchers have studied ways to detect and ignore such manipulated rankings, and rival methods such as HITS, TrustRank and SALSA approached the same problem from other angles. Earlier still, Robin Li's RankDex search engine, launched in 1996, had ranked sites by how many others linked to them; Page later cited Li's work in his patents.
What PageRank teaches about judging anything
PageRank is a formal version of an everyday shortcut: we trust what trusted people point to. That shortcut is powerful because it pools the judgment of many independent people, and fragile for the same reason, since it rewards whatever gets pointed at, which is not always the same as whatever is true.
It also shows how advantages compound. A page with many good links becomes easier to find, so more people link to it, which raises its score again. Anyone using a ranked list, from search results to bestseller charts, is looking at the output of a loop like this, and is wise to ask how much of the ranking reflects quality and how much reflects an early head start.
Questions people ask
Does Google still use PageRank?
No longer on its own. PageRank was the first algorithm Google used to order search results and is still the best known, but it is one factor among many in today's ranking. The patents associated with PageRank all expired in September 2019, and the score is no longer the deciding signal it once was.
Why is it called PageRank?
The name works on two levels: it ranks web pages, and it carries the surname of its co-inventor, Larry Page. The word is a Google trademark, while the original patent was assigned to Stanford University, which licensed it exclusively to Google in exchange for shares in the company.
What is the damping factor in PageRank?
It is the probability that the imaginary random surfer keeps clicking links instead of jumping to a random page. It is commonly set around 0.85. The damping factor stops loops of pages from hoarding score and makes sure every page keeps a small minimum value, even one that nothing links to.
The thinking behind it
John MacCormick gives PageRank a full chapter, with the random-surfer idea worked through in plain language alongside eight other algorithms that shaped computing.
Read or listen to Nine Algorithms That Changed the Future
Hear the whole book free: start an Audible trial and your first audiobook — this one, if you like — is on the house.
As an Amazon Associate, ReadGlobe earns from qualifying purchases and Audible trials — at no extra cost to you.
Sources
- PageRank — Wikipedia
- Link analysis — Wikipedia
How this was made: Amili, an AI writer, wrote this article in its own words from the sources above. Every link was checked before publishing. Spotted an error? Tell us and we will correct it.