PageRank

Definition of PageRank

PageRank is an algorithm invented in 1998 by Larry Page and Sergey Brin, then PhD students at Stanford. It serves as the foundation of Google, the search engine they launched the same year. The idea fits in one sentence: the value of a web page can be estimated by looking at the number and quality of the links that point to it.

That intuition was revolutionary. Before Google, search engines ranked pages mostly from their textual content. AltaVista, Yahoo, Lycos compared the query keywords to the words present in the page. PageRank added a new dimension: a page's reputation, measured by the implicit votes that inbound links represent.

How it works

The principle mirrors that of an academic citation. A scientific paper frequently cited by other important papers gains authority in its field. In the same way, a web page that receives links from other already-recognised pages sees its score climb.

The mechanic is recursive. The PageRank of a page A depends on the PageRank of the pages B, C, D that point to it. But the PageRank of B itself depends on the pages that point to B, and so on. To solve that multi-variable equation, the algorithm proceeds through successive iterations, recomputing scores until they converge to a stable value.

Another important subtlety: the PageRank of a page is distributed among all the pages it links to. If a very powerful page emits twenty links, each link transmits one twentieth of its juice. That property explains why it is better to receive a link from a page that emits few than from one that emits hundreds.

The original formula

The formula published by Page and Brin in their 1998 paper uses a damping factor of 0.85, which models the probability that an internet user clicks on a link rather than typing a new URL at random. That factor prevents the calculation from diverging and reflects a behavioural reality: users do not chain links indefinitely.

The algorithm walks the entire page graph, propagates PageRank step by step, and generally converges after about thirty iterations. On the original Google scale, scores ranged from 0 to 10 on a logarithmic basis. Moving from PageRank 4 to 5 took far more effort than moving from 1 to 2.

That formula has been published and widely studied since. It remains a study object in graph theory and has applications well beyond the web: social-network analysis, product recommendation, ranking of scientific articles.

Evolution since 2000

Google has continually enriched its algorithm since the original PageRank. Hundreds of signals have been added, some major: content quality, user experience, behavioural signals, search intent, mobile-friendliness, speed, HTTPS security. Each major algorithm update (Penguin, Panda, BERT, Helpful Content) has reshuffled the weightings.

PageRank itself has been refined. More recent versions weight links based on their quality, their position on the page, their semantic context. A link in the body of a relevant article no longer carries the same weight as a link in a generic footer. The nofollow tag, introduced in 2005, allowed sites to indicate that a link should not pass PageRank.

In 2016, Google definitively removed the PageRank Toolbar, the public score visible in the Google bar of browsers. That removal did not signal the disappearance of the algorithm, only its closure to outside observers. Google explained that the public score encouraged manipulation behaviour and no longer offered a useful reading at scale.

PageRank in 2026

Twenty-five years after its creation, PageRank remains active at the core of Google's algorithm. John Mueller and several Google engineers have repeatedly confirmed that links and their authority propagation still rank among the heaviest signals. What has changed is the environment.

Today, PageRank coexists with hundreds of other signals. A page can rank well on a query without an exceptional PageRank, if its content matches search intent perfectly and offers an exemplary user experience. Conversely, a high PageRank alone is no longer enough to compensate for poor content or a slow page.

Algorithmic detectors are also more sophisticated. Google identifies manipulative link profiles, link farms and low-cost PBNs. PageRank earned through these routes is devalued, sometimes penalised. That evolution favours editorial, contextual and therefore more natural link acquisition.

Practical implications for SEO

For an SEO or an advertiser in 2026, several concrete implications flow from PageRank. First, internal linking remains an under-used lever. PageRank flows between your pages through internal links: structuring that linking as a topical cluster, channelling juice toward priority pages, is optimising a signal that was already valued back in 1998.

Second, quality trumps quantity in a netlinking strategy. Receiving a link from a page recognised in its topic is worth far more than ten links from pages without an audience, which is the rationale behind buying a backlink on a high-authority media rather than spreading budget across dozens of weaker placements. Source selection should beat volume.

Third, the transparency of anchors matters. PageRank is sensitive to link context: coherent anchors, position in the body of the article, relevant editorial environment. Links that look like natural placements pass through better than those that scream their commercial nature, which is why teams that publish on media that genuinely pass authority tend to invest in the editorial fit before negotiating the placement itself.