Module 6: PageRank & Link Authority
30 minutes
PageRank: The Original Signal
PageRank (named after Larry Page) models the web as a directed graph where links are votes. A page's PageRank is a function of:
- The number of pages linking to it
- The PageRank of those linking pages
- The number of other links on each linking page (dilution)
The random surfer model: Imagine a user who randomly clicks links. The probability they end up on any given page, after many random clicks, is its PageRank. Pages at the center of the link graph — the "hubs" — are most likely to be visited.
Why it worked: In 1998, links were primarily editorial — someone chose to link because they found something valuable. PageRank was measuring genuine recommendation signals.
The Reasonable Surfer: PageRank Evolved
The original PageRank model had a problem: it treated all links on a page as equally likely to be clicked. A link in the body of an article was worth the same as a link in the footer boilerplate.
The Reasonable Surfer patent (2004-2010) fixed this by modeling realistic click probability for each link. Links with high click probability (in-content, descriptive anchor, prominently placed) pass more PageRank. Links with low click probability (footer, "click here," hidden) pass almost none.
The Reasonable Surfer factors (see Reasonable Surfer Link Audit):
- Position in the page (above fold main content = highest)
- Visual prominence (size, contrast, styling)
- Anchor text quality (descriptive vs. generic)
- Context relevance (surrounding text supports the link)
- Link type (editorial in-content = highest; boilerplate = lowest)
Internal linking implication: Placement of internal links matters as much as their existence. A footer link to your pricing page passes almost no PageRank. An in-content editorial mention of pricing in a relevant blog post passes significant PageRank.
Agent Rank: Author Authority
The Agent Rank patent extended the authority model from pages to authors. The core idea: trust an author based on their track record, not just the site they publish on.
The mechanism:
- An author's content across the web builds an "Agent Rank" — an authority score for specific topic areas
- When that author publishes a new document, some of their Agent Rank flows to that document
- A high-trust author publishing on a low-authority domain can outrank a low-trust author on a high-authority domain
Why this matters: The E-E-A-T framework — Experience, Expertise, Authoritativeness, Trustworthiness — is in part an algorithmic implementation of Agent Rank. Author signals are ranking signals.
The Agent Rank Author Audit operationalizes this: build the byline signals, author pages, credentials, and cross-platform presence that feed the author authority model.
Link Manipulation and the Historical Data Patent
The Historical Data patent (by Matt Cutts and Paul Haahr — the engineers who built the link spam detection systems) explains how Google detects manipulated link profiles through temporal analysis.
The key insight: natural link acquisition has characteristic temporal patterns. Manipulated link profiles have different patterns. The algorithm learns the difference.
Natural link velocity: Gradual organic growth, variance correlated with content publication, peaks tied to real-world events
Manipulated link velocity: Sudden spikes with no correlating content/news event, burst-then-stop pattern, cluster acquisition with uniform anchor text
The Historical Data Risk Analyzer translates this into an auditable checklist.