PageRank

On this page 7

PageRank is a link analysis algorithm that estimates the importance of a web page from the quantity and quality of the links pointing to it, treating each link as a weighted vote passed from one page to another. Developed by Larry Page and Sergey Brin at Stanford University in the late 1990s, it became the foundation of Google Search and remains a central idea in SEO, particularly in the field of links & authority.

PageRank is no longer the only or dominant ranking factor, and Google stopped showing a public score in its toolbar in 2016. The underlying principle, that links transfer authority in proportion to the importance of the linking page, still shapes how search engines evaluate the web and how practitioners think about link equity.

How the PageRank algorithm calculates importance

PageRank calculates importance by modeling a random surfer who clicks links at random and occasionally jumps to an arbitrary page. The PageRank of a page is the probability that the surfer is on that page at any given moment after browsing for a very long time. Pages that many important pages link to accumulate a higher probability.

In the original formula, the score of page A equals a baseline term plus a damping factor d multiplied by the sum of contributions from every page linking to A. Each linking page contributes its own PageRank divided by its number of outbound links, so a link from a page with few links passes more value than a link from a page with hundreds. The calculation is repeated iteratively until the scores converge, which mathematically corresponds to finding the principal eigenvector of the web’s link matrix.

The table below describes the main parameters of the classic model.

ParameterTypical value or formEffect on the result
Damping factor (d)About 0.85Share of value passed through links on each step
Teleportation probability1 minus d, about 0.15Prevents value from being trapped in closed loops
Outbound link countNumber of links on the source pageDivides the value each link passes
IterationsOften 50 to 100 on large graphsRepeated until scores stabilize
Dangling nodesPages with no outbound linksTheir value is redistributed across the graph
Parameters of the classic PageRank model

Modern search engines use refined descendants of PageRank rather than the 1998 formula. The original model treated every link on a page equally, which made it easy to manipulate through link farms and paid links. Later patents and public statements describe several refinements that make link value depend on context.

One documented refinement is the reasonable surfer model, which assigns more weight to links that a real user is likely to click, such as a prominent link in the main content, and less to links in footers or boilerplate. Other systems consider topical relevance between the linking and linked pages, the trustworthiness of the source and whether a link carries attributes such as nofollow, sponsored or ugc.

StageMain ideaWeakness addressed
Classic PageRankEvery link is an equal votePure keyword matching ignored authority
Reasonable surferLink weight depends on likely clicksFooter and sitewide links were overvalued
Trust-based seedingValue starts from vetted sitesSpam networks generated artificial authority
Link attributesNofollow, sponsored and ugc as hintsPaid and user-generated links distorted scores
Spam classifiersMachine learning discounts manipulative linksLarge-scale link schemes and link farms
Evolution of link-based importance signals

PageRank is an internal search engine signal, whereas metrics such as Domain Authority and Domain Rating are third-party estimates built by SEO tool vendors from their own link indexes. They are inspired by the same idea, but they are not used by Google and they cannot be converted into one another.

MetricOwnerLevel measuredVisibility
PageRankGoogleIndividual pageInternal only since 2016
Domain AuthorityMozWhole domainPublic score from 1 to 100
Domain RatingAhrefsWhole domainPublic score from 0 to 100
Link equityIndustry conceptValue passed by one linkNot a measured score
PageRank and similar link metrics

A key distinction is that PageRank is calculated per page, not per domain. A strong homepage does not automatically make every deep page strong; authority reaches those pages only through internal links. This is why site architecture and internal linking influence how link value is distributed across a website.

Using the PageRank principle in SEO strategy

Practitioners apply the PageRank principle by earning links from important, relevant pages and by directing internal link value toward the pages that matter commercially. Because the real score is invisible, the work relies on proxies and on a sound understanding of how value flows.

SituationRecommended actionReason
Key page buried deepLink to it from strong hub pagesInternal links pass accumulated value
Many low-value outbound linksKeep navigation focusedEach extra link dilutes the value per link
Broken page with backlinksApply a 301 redirect to a relevant pageRecovers value that would otherwise be lost
Orphan pagesAdd contextual internal linksPages without links receive little value
Paid placementsMark links as sponsoredAvoids violating search engine spam policies
Practical decisions shaped by PageRank

PageRank sculpting, the old practice of adding nofollow to internal links to concentrate value, no longer works as intended. Since 2009, Google has indicated that value assigned to a nofollowed link is not redistributed to the other links on the page; it is simply lost. Better results come from a clear site hierarchy with fewer, more meaningful links.

Misconceptions and limitations

PageRank is frequently misunderstood because its public toolbar version was a coarse 0 to 10 logarithmic scale that showed only a rounded snapshot of a much more detailed internal value. Several persistent beliefs come from that era.

  • PageRank is not dead. Google has confirmed that link-based importance is still used, even though the public score disappeared.
  • PageRank is not the ranking. It is one of hundreds of signals, alongside relevance, content quality and user experience.
  • More links are not always better. Links from link farms, private blog networks or irrelevant sites are discounted or can trigger penalties.
  • A high third-party score is not PageRank. Tool metrics only approximate link strength from incomplete link indexes.

The algorithm also has structural limits. It measures popularity in the link graph, not accuracy or usefulness, so new pages with excellent content may rank poorly until they attract links. This gap is one reason why search engines combine link signals with content and quality evaluation rather than relying on links alone.

Understanding it simply

Imagine the web as a huge town where every link is a recommendation. If many people recommend a bakery, it is probably good. If the town’s most respected food critic recommends it, that single recommendation counts even more.

PageRank turns this idea into math. A page becomes important when important pages link to it, and a page that recommends hundreds of others gives each of them only a small share of its credibility. Search engines still use this logic today, combined with many other checks on quality and relevance.

Applied in practice

An online retailer of hiking gear noticed that its new guide to choosing trekking poles received almost no search traffic, even though the content was thorough. An audit showed that the guide sat five clicks from the homepage and had only one internal link, from an archive page.

Meanwhile, the homepage and a popular backpack buying guide had attracted more than 400 backlinks from outdoor blogs and regional news sites. The team added a contextual link to the trekking pole guide from the backpack guide, featured it in the main navigation under Buying advice and redirected two outdated pole pages, which still had external links, to the new guide with 301 redirects.

Within about two months, the guide moved from page three to the top five results for its main queries. No new backlinks were acquired; the improvement came from redistributing existing link value through internal links, exactly as the PageRank model predicts.