Understanding the PageRank patent

This is Larry Page's original PageRank patent — "Method for node ranking in a linked database," filed by Stanford in January 1998, issued to Google in September 2001.

What it covers

The core idea: assign an importance score to each document in a linked collection (like the web) based on how many other documents link to it — and weighted by how important those linking documents are themselves. A link from a highly-ranked page counts more than a link from an obscure one, so importance flows recursively through the citation/link graph.

Key mechanisms in the claims:

• Basic scoring (claim 1): score a document from the scores of documents linking to it.
• Weighting factors (claims 2–7): the influence of a linking document can depend on things like how many outbound links it has, the probability a user actually visits it, its URL/domain/author/update time, whether it's a "root" document, or how prominently the link is displayed.
• Random walk model (claims 9–11): simulates a user randomly clicking links (with some probability of jumping to a random page instead) — the more often a document gets visited in this simulation, the higher its rank. This is the classic "random surfer" formulation of PageRank.
• Output/use (claims 12–17): the ranks can be shown as directory listings, annotated with icons/bars/text, combined with text matching (including anchor text), or used to group documents.
• Computer-readable medium claims (18–19): the same method as an instruction set stored on media.

Status: Expired (lifetime) — filed 1998, anticipated expiry 2018, now long out of force. Currently assigned to Google LLC and Stanford, originally solely to Stanford (Stanford licensed it to Google, which was standard for university-originated tech in that era).

This patent is foundational — it's essentially the mathematical/algorithmic basis that became Google Search's original ranking engine.