PageRank is a Markov Chain Model
Looking at PageRank as a Markov chain can change how you understand PageRank.
A Markov chain is a mathematical way of modeling what happens when movement from one state to another is based on probability. Applied to the web, the states are pages and the links between them provide the paths a hypothetical random surfer could take.
PageRank uses the link graph to model how importance could be distributed across the web. By modeling how a reasonable surfer might follow those links, the system eventually arrives at a stable distribution. That distribution produces a PageRank score for each page, representing its relative importance within the link graph.
In this sense, PageRank is a mathematical prediction rather than a measurement of what people are actually doing. It predicts how importance would be distributed if users followed the link graph according to the model.
This distinction becomes important when considering how modern search systems may go beyond the original PageRank model. AI systems can work with probability distributions and can potentially combine link-based predictions with other sources of information, including observations of which pages real people actually visit.
Exactly how Google combines PageRank with other signals is proprietary and has not been fully disclosed.
Random surfer vs. reasonable surfer: Google patent US7716225B1 is associated with the shift from the random-surfer concept toward a reasonable-surfer model, in which different links can have different probabilities of being followed. A patent protects a method rather than a single application of that method. For example, a patented gear can be used in different applications without each application having to be individually specified in the patent. The reasonable-surfer concept therefore provides a basis for considering whether a link is likely to be followed rather than simply treating every link as an equal vote.
The Strategic Error of Considering PageRank a Popular Vote
Although PageRank can appear to produce an emergent popularity or voting factor for content, viewing it as a popular vote overlooks its actual function as a probability-based model of how importance is distributed across the web's link graph.
Viewing a link as a vote leads to the belief that a footer link transfers importance in the same way as a link within the contextual content of a page, and that a reasonable surfer is equally likely to follow either link. Google's reasonable-surfer improvements move beyond treating every link as an equal indication of importance. A footer link may contribute to the discovery and crawling of a page without producing the same user-engagement signals as a link that people actually follow. NavBoost may then override or refine the prediction based on evidence of how users actually engage with particular links.
This distinction also undermines strategies based simply on increasing the quantity of links pointing to a site, or creating numerous nearly identical category pages with the same set of links pointing to important pages. A link that is hidden or unlikely to be seen or followed by users may provide little or no value as an indication of importance.
The "links as votes" model also ignores the predictive nature of PageRank. If an algorithm determines that a link has no semantic relevance, represents a bait-and-switch, or does not match user intent, the link may lose its importance or potentially become a negative signal. Likewise, if near-duplicate pages all contain the same links, those pages may be purged from the index or their links discounted because they provide no additional predictive value of importance.
Important Links Strategy
An important-links strategy does not mean eliminating footer links. A visitor who lands on a page may read the content and then visit the About page. In that case, the About page is an important page to that visitor, and systems such as NavBoost can provide evidence of that importance through observed user behavior.
Links to consensus URLs are important for visitors who may be uncertain whether an assertion is correct. While many people consider PageRank a voting system and attempt to hoard PageRank, adding outbound links to a page does not reduce that page's PageRank. The global damping factor prevents PageRank from being trapped within a site, so outbound links do not simply cause PageRank to leak away from the site.
There is, however, a possibility that a visitor will leave the site through an outbound link and not return. The value of providing supporting consensus information can be more important than the possibility of losing a visitor. Some visitors may simply return to search instead. A visitor who finds useful information on your page and then follows a supporting link may be more valuable than a visitor who returns to search because the page did not satisfy their need.
The AI-Aware SEO Strategy: Consider links based on the value they provide rather than simply treating them as votes. For backlinks, the Search-Everywhere concept asks whether the link can make the site discoverable to people who may be looking for its information or solutions. Is the backlink likely to produce relevant traffic?
For on-page SEO, the question is whether the link helps match the visitor with the intent that brought them to the page. A useful link can help visitors discover additional information, verify an assertion, or continue toward a solution—even when that link takes them to another site.
Published
by Wayne Smith – The SEO Heretic
Wayne Smith has worked in online marketing, search, and web development for several decades. The role of an SEO heretic is to question dogma and tradition when empirical evidence shows it is incorrect.