PageRank

Long definition

PageRank is Google's founding algorithm. Larry Page filed the original patent (US 6,285,999) at Stanford in January 1998, and Brin and Page co-authored the seminal paper "The Anatomy of a Large-Scale Hypertextual Web Search Engine" the same year. The algorithm models the web as a directed graph and computes each page's rank iteratively: a page's score is the sum of the scores of pages linking to it, divided by their respective outbound link counts, plus a damping factor.

The intuition: a link is an editorial vote, votes from authoritative sites count more than votes from random sites, and authority itself is determined by who else votes for you. Convergence over many iterations produces stable rank values for every page in the graph. The damping factor (typically 0.85) models a "random surfer" who occasionally jumps to an arbitrary page rather than following links forever.

Public exposure of PageRank evolved and ended:

• 2000-2016 — Toolbar PageRank: a 0-10 logarithmic score visible in the Google Toolbar browser plugin. Heavily gamed by SEOs and link sellers.
• March 2016 — Google formally retired Toolbar PageRank and removed the public-facing score.
• 2016-present — PageRank remains internal. Google's Gary Illyes confirmed in 2017 that "Google still uses PageRank — internally, just like 18 years ago." The 2024 Content Warehouse API leak revealed numerous PageRank-related fields still in active use.

Modern PageRank in Google's stack is layered on top of many other signals. The original algorithm assumed all links are equal in nature; modern Google distinguishes link types (nofollow, sponsored, ugc), spam-pattern devalues, applies topical relevance, and combines PageRank flow with neural ranking signals. The conceptual currency that connects all of this is what we now call link equity.

The patent itself is now expired (as of 2018, given its 1998 priority date), but the algorithm has been so heavily evolved that the patent text describes only an early state of what the system actually does.

Common misconceptions

• "Google killed PageRank." Google killed the public score. The internal algorithm remains a foundational ranking signal — confirmed by Google staff multiple times since 2016.
• "PageRank is just about link count." Quantity matters, but quality and graph topology matter more. A single link from a highly-ranked source can move PageRank more than thousands of links from low-ranked sources. The recursive nature of the algorithm makes context essential.
• "Toolbar PageRank reflected real ranking influence." It was a snapshot of internal PageRank, exported infrequently, often months out of date, and subject to manipulation by aggressive linking. Real ranking influence today involves dozens of signals beyond PageRank.
• "PageRank sculpting via nofollow still works." It hasn't worked since 2009. Google distributes PageRank across all outbound links regardless of rel attributes; nofollow just means the share for that link doesn't pass — it isn't redirected to the followed links.