PageRank

PageRank is an algorithm that scores how important a page is based on the links pointing to it. It was described by Larry Page and Sergey Brin at Stanford in 1998 and became one of the foundations of Google Search. The core idea is that a link is a vote, and votes from important pages count for more than votes from obscure ones.

How it works

Imagine a “random surfer” who clicks links at random and occasionally jumps to a random page. A page’s PageRank is the probability that the surfer is on it at any moment. Each page shares its score across the pages it links to, so a page linked from many well-linked pages accumulates a high score. A damping factor (classically 0.85) models the chance of jumping to a random page instead of following a link. The scores are calculated by iterating over the whole link graph until they settle.

Google stopped showing public Toolbar PageRank scores in 2016, but has said that PageRank, in evolved forms, is still one of many signals in its ranking systems. The same maths is widely used outside search as a measure of how central a node is in any graph, including entity and knowledge graphs.

How it’s used in SEO and marketing

• Internal linking: the idea of “PageRank flow” explains why pages buried deep in a site, or linked only from weak pages, get less attention. MLforSEO’s customer segmentation practical notes that a logical structure gives better internal PageRank flow.
• Internal link audits: you can compute an internal PageRank for your own site from a crawl (for example with Python’s networkx) to find important pages that receive too few links.
• Link building strategy: one relevant link from a strong, topically related page is worth more than many weak ones.
• Graph analysis of entities and topics: rank which entities or topics are most central in a co-occurrence network.

Example

A crawl of a 2,000-page site is loaded into networkx and PageRank is calculated on the internal link graph. The top pages are the homepage, category hubs and two old blog posts that the main navigation still links to. A key product guide sits near the bottom, which leads the team to link it from the relevant hubs.

Limitations and things to watch

• PageRank alone says nothing about relevance; it measures link-based importance only.
• Internal PageRank calculations are a model of your site, not a view of Google’s actual scores.
• In AI search, MLforSEO’s Agentic SEO course points out that authority is increasingly claim-specific: a high-authority domain can still have individual claims excluded if other sources do not support them. Links help discovery and trust, but they do not replace precise, verifiable content.

Learn it on MLforSEO

• AI Search Optimisation and Agentic SEO, lesson on why traditional SEO is not enough
• 10 Ways to Automate Internal Linking with AI
• Blog and Site Architecture for SEO and Conversions in the AI Search Era
• Introduction to Machine Learning for SEO