reasonable surfer model

Google’s original PageRank assumed a “random surfer” who was equally likely to click any link on a page. The reasonable surfer patent (inventors Jeffrey Dean, Corin Anderson, Alexis Battle) drops that assumption. It models a user who follows some links more than others, and it lets the weight a link passes scale with that click probability.

What changes a link’s weight

The patent names concrete features that move the number: anchor text font size and color, position on the page (in-body versus a sidebar, footer, or a long list), the number of words in the anchor, how commercial the surrounding language reads, topical relevance between the two pages, and whether the link stays on the same domain. It also references user-behavior data, meaning actual clicks on a given link.

Example: an in-content link inside a relevant paragraph, high on the page, with descriptive anchor text is treated as more likely to be clicked than a “Terms of Service” link in the footer or a banner ad. Under this model the first link can pass meaningfully more equity than the second.

Two cautions. A patent shows what Google explored, not a confirmed production algorithm, and this original filing has since expired (a continuation, US8117209B1, followed). Even so, the behavior it describes, that prominent in-context links matter more than buried sitewide ones, lines up with what practitioners observe and with Google’s own guidance on link placement.