PageRank explained: does Google still use it?
PageRank is Google's original system for scoring pages by the links that point to them. Google still uses it. Its ranking systems guide lists PageRank, and a 2025 court exhibit describes the modern version as distance from a known good source.
In short
- PageRank was published by Sergey Brin and Larry Page in 1998 and scores a page by the chance that a random surfer who keeps clicking links lands on it.
- In the original formula each page divides its score among its outgoing links, with a damping factor that the authors usually set to 0.85.
- Google's ranking systems guide, last updated in December 2025, still lists link analysis systems and PageRank among its ranking systems.
- A Google engineer interview published as a court exhibit in May 2025 describes PageRank as a single signal relating to distance from a known good source, used as an input to the quality score.
- The public 0 to 10 Toolbar PageRank score is reported to have stopped updating in 2013 and to have been removed in 2016, so no tool shows real PageRank today.
PageRank is the system Google was founded on. It scores a page by the links that point to it, and it treats a link from a well-linked page as worth more than a link from an obscure one. Google still uses it. The public score disappeared years ago, and the version running today is described differently from the 1998 original, but links still feed a PageRank signal inside Google’s ranking systems. For anyone learning about links this is the idea underneath almost everything else, because most of the advice you will read about which links are worth having is an attempt to guess what PageRank, in its current form, rewards.
The original idea
Sergey Brin and Larry Page described PageRank in their 1998 paper on the Google prototype, “The Anatomy of a Large-Scale Hypertextual Web Search Engine”. They explained it with a thought experiment. Imagine a “random surfer” who starts on a random page and keeps clicking links, never going back, until they get bored and start again on another random page. The probability that the surfer visits a page is that page’s PageRank. The picture is useful because it needs no mathematics to follow. A page with many routes leading to it will be reached often, a page that nothing links to will be reached only when the surfer happens to start there, and a page linked from busy pages will be busier than one linked from quiet corners.
Two consequences follow, and both still shape how people think about links. The first is that a page shares its score among its links. The paper says the score is normalised by the number of links on the page, so a page with ten outgoing links gives each a smaller share than a page with two. The surfer can only click one link at a time, and the more there are to choose from, the less likely any one of them is to be chosen. The second consequence is that scores pass along chains. A page is strong if strong pages link to it, and those pages are strong for the same reason. The paper also includes a damping factor, the chance that the surfer keeps clicking instead of jumping away. The authors wrote that they usually set it to 0.85, which leaves a small chance at every step that the surfer gives up and starts somewhere new, and that in turn means value fades a little each time it passes along a chain.
The formula in plain words
The 1998 paper gives this formula for a page A that is linked from pages T1 to Tn:
PR(A) = (1 - d) + d × (PR(T1)/C(T1) + ... + PR(Tn)/C(Tn))
Here d is the damping factor and C(T) is the number of links going out of page T. In words: take each page that links to A, divide its PageRank by the number of links it has, add those shares together, and scale the total by the damping factor. A simple case shows the effect. Suppose two pages link to you, and both have the same score. One has 5 outgoing links and the other has 50. The first passes you one fifth of its score. The second passes you one fiftieth. That is the origin of the advice to prefer pages that link out sparingly, and of the idea of link equity, which is the trade’s name for the value a link hands on.
The same paper introduced a second use of links. It says anchors “often provide more accurate descriptions of Web pages than the pages themselves”. Anchors are the clickable words of a link, and the observation is that what other people call a page is often a better guide to its subject than what the page says about itself. That is why anchor text became a ranking signal alongside PageRank. So from the very first paper a link was doing two jobs at once, with PageRank measuring how much a page deserved attention and the anchor describing what the page was about, and the two have stayed separate signals ever since.
What happened to the public score
For years Google showed a 0 to 10 PageRank score in its browser toolbar. An industry grew up around it, including the sale of links priced by the score of the page. Once a number is visible and can be bought, people stop treating it as a measurement and start treating it as a target, and the link market that resulted worked directly against what the score was meant to measure. The public score is reported to have stopped updating in 2013 and to have been removed in 2016.
Since then, nobody outside Google can see a page’s PageRank. The scores in SEO tools, such as Domain Rating, URL Rating and Domain Authority, are each vendor’s own estimate built on its own crawl. They borrow the idea, and none of them is a Google number. This is one of the most common confusions among beginners, who see an authority score in a tool and assume Google supplied it or at least recognises it. Each vendor crawls its own portion of the web and applies its own formula, which is why two tools can give the same site quite different scores. Google has denied using Domain Authority. Domain Authority vs Domain Rating explains what each one measures.
The evidence that Google still uses it
Three sources agree, and they are worth taking in order of how much weight they can bear. The strongest is Google’s documentation. The ranking systems guide, last updated in December 2025, lists “Link analysis systems and PageRank”. It says Google has various systems that understand how pages link to each other, that PageRank is one of the core systems used since Google launched, and that it has evolved a great deal since the original paper. The second source is court evidence. A Google engineer interview, published as an exhibit in the remedies phase of the US antitrust case and reported in May 2025, says: “PageRank. This is a single signal relating to distance from a known good source, and it is used as an input to the Quality score.” That is a statement made in legal proceedings, in an engineer’s own words, and it is the clearest recent description available.
The third source is the 2024 leak, and it is the one to treat with most care. The internal documentation that leaked in May 2024 contains several PageRank fields, according to iPullRank’s analysis. They include one whose name analysts read as “nearest seeds”, older variants marked as deprecated, a toolbar PageRank field, and a field described as the PageRank of the site’s homepage, attached to every page. The leak lists fields. It does not show weights or prove that a field is in use. A field is a labelled slot where a system can store a value, so its presence shows that Google built somewhere to keep the number and nothing about how the number is used today. Our lesson on the leak and court documents covers the limits.
How today’s PageRank differs
The core idea has changed. In 1998 PageRank was the probability that a random surfer lands on the page, and every page counted as a possible start. What is known today, per the 2025 court exhibit, is that the signal relates to distance from a known good source, so the starting points are known good sources. Which sources those are is not public. The seed idea matters in practice. If the signal measures how close a page is to trusted sources, then a link that sits one or two steps from well-established sites is worth more than a link deep inside a cluster of sites that only link to each other. Under the old model such a cluster could build up a score among its own members, whereas a model that starts from trusted sources gives it nothing unless trust flows in from outside. Lists of “seed sites” that circulate online are guesses.
The role in ranking has changed too. The original PageRank was the main quality measure, and the public could see it as a toolbar score from 0 to 10, where today there is no public score at all. It is now one input to a quality score that is described as largely the same across queries. The court exhibit also places PageRank on the quality side of ranking, not the relevance side. Relevance to a query comes from anchors, page text and clicks. PageRank contributes to whether the site is trusted at all. Put simply, PageRank helps Google decide whether your site belongs among the candidates, and other signals decide which candidate best fits a particular search. How search engines evaluate links sets out both sides.
The treatment of nofollow links is the last difference. Nofollow is a label a site can put on a link to ask search engines not to pass value through it, and the attribute did not exist until 2005, so the 1998 model has every link passing a share. In June 2009, Google’s Matt Cutts explained a change. If a page has ten links and five are nofollowed, the five normal links each pass a tenth of the page’s PageRank, not a fifth. The share of the nofollowed links is not handed to the others, so a nofollowed link still uses up its share. That ended the practice of “PageRank sculpting”, where site owners nofollowed some links to push more value through the rest. Since 2019 nofollow is a hint, which means Google may choose to count such a link after all. Link attributes covers how nofollow, sponsored and ugc work now.
What this means for your links
The first practical lesson is to chase the source and not a score. A tool’s authority score is an estimate. Whether real, trusted sites link to the page that links to you is closer to what PageRank measures, so when you assess a possible link it is worth looking at who links to that site and whether those are publications you recognise. The second lesson is to expect each extra link from the same site to add less, because the model rewards new sources. A tenth link from a site that already links to you nine times opens no new route for the surfer, while a first link from a site that has never mentioned you does.
It also pays to look at the linking page itself. A page with hundreds of outgoing links divides whatever it has between all of them, which is the arithmetic of the formula applied to a real decision, and it is why a link on a long list of “resources” is usually worth less than a link in an article that cites only a few sources. Finally, do not expect PageRank alone to rank a page. Google’s Gary Illyes said in April 2024 that Google needs “very few links” to rank pages. Do backlinks still matter puts the signal in proportion. Links that journalists and editors give freely tend to come from the well-connected sites this model favours, which is one reason digital PR is rated so highly.
Where to go next
The next lesson covers how value moves through redirects and internal links, which takes the sharing rule described above and follows it through the practical cases a site owner meets, such as a page that has moved to a new address. To compare the tools that estimate link strength, see backlink analysis tools, keeping in mind that every score they show is the vendor’s estimate and none is PageRank. Definitions are in the glossary, including the terms used in this lesson that later lessons will take for granted, such as anchor text, nofollow and referring domain.
Common questions
Does Google still use PageRank?
Yes. Google's ranking systems guide lists link analysis systems and PageRank, and says PageRank has evolved a great deal since the original paper. A 2025 court exhibit describes it as an input to Google's quality score.
Can I check the PageRank of my site?
No. The public Toolbar PageRank score is reported to have been removed in 2016. Domain Rating, Domain Authority and similar scores are estimates made by tool vendors from their own crawls. None of them is PageRank.
What is the damping factor in PageRank?
It is the chance that the imaginary surfer keeps clicking links instead of jumping to a random page. Brin and Page wrote in 1998 that they usually set it to 0.85.
Do nofollow links pass PageRank?
A nofollow link is a request not to pass it. Matt Cutts of Google wrote in June 2009 that a nofollowed link still uses up its share of the page's PageRank. Since 2019 Google treats nofollow as a hint, so it may choose to count a link.
Is PageRank the same as Domain Authority?
No. PageRank is Google's internal signal. Domain Authority is a Moz metric and Domain Rating is an Ahrefs metric. Google has denied using Domain Authority.

