Ranking and link analysis: patents and papers
Ranking and link-analysis patents and papers describe how a search engine can score pages from the links between them, from how people use those links, and from click behaviour. The best documented is PageRank: Google calls it one of its core ranking systems at launch and links to its original paper and patent. Nine patents and seven papers are listed, all verified.
What does this theme cover?
Google says it has “various systems that understand how pages link to each other” and that PageRank was “one of our core ranking systems used when Google first launched”, adding that how PageRank works “has evolved a lot”.[1] Officially documented The other documents here describe refinements: weighting links by how likely they are to be followed, measuring distance from trusted seed pages, using a document’s history, and learning from clicks. For practice, see links and link building.
Which patents describe PageRank and link-based ranking?
Method for node ranking in a linked database
What it describes: Ranking nodes in a linked database (such as the web) where a document’s rank is calculated from the ranks of the documents citing it, plus a constant representing the probability that a browser will randomly jump to the document.
Why it matters for SEO: This is the PageRank patent. Google links to it from its ranking systems guide,[1] which makes it the rare patent tied to a confirmed system. It explains why links from important pages count more than links from obscure ones. The patent is listed as expired.
Method for scoring documents in a linked database
What it describes: A later patent in the PageRank family (same January 10, 1997 priority date) on scoring documents in a linked database from the documents that link to them.
Why it matters for SEO: Part of the same patent family as US 6,285,999. Useful if you are tracing how the original PageRank claims were extended.
Method for node ranking in a linked database
What it describes: Another later patent in the PageRank family (same 1997 priority date) on node ranking in a linked database.
Why it matters for SEO: Confirms the family lineage; no new SEO takeaway beyond PageRank itself.
Ranking documents based on user behavior and/or feature data
What it describes: A model built from features of a link (the description lists examples such as the font size of the anchor text and the position of the link on the page) and from user behaviour data about which links people follow; the model is used to rank documents.
Why it matters for SEO: Often called the “reasonable surfer” patent. It suggests not all links pass equal value: a prominent, in-content link that people are likely to click may count more than a buried footer link. Earn editorial links in the body of relevant pages. Practitioner practice
Ranking documents based on user behavior and/or feature data
What it describes: A later patent in the reasonable-surfer family (same June 17, 2004 priority date) with the same title and abstract.
Why it matters for SEO: Shows Google kept pursuing claims on this family into 2016. Same practical takeaway as above.
Producing a ranking for pages using distances in a web-link graph
What it describes: Ranking pages by their distance from a set of seed pages in the link graph. Links are assigned lengths based on properties of the links and pages, and shortest distances from the seeds determine the ranking scores.
Why it matters for SEO: Implies that links close (in link hops) to trusted, well-linked sites matter more than links from isolated corners of the web. Prioritise links and citations from reputable sources in your field. Practitioner practice
Information retrieval based on historical data
What it describes: Scoring a document using one or more types of history data associated with it (for example when it first appeared or how it and its links change over time).
Why it matters for SEO: Freshness and link history are plausible signals. Keep important pages maintained and dated honestly; see AEO.wiki on freshness.
Document scoring based on document inception date
What it describes: Scoring and ranking a document based partly on its inception date (when the document is first known to the system).
Why it matters for SEO: Part of the same historical-data family. Changing dates without changing content does not change when a page was first seen.
Modifying search result ranking based on implicit user feedback
What it describes: A relevance measure for a result, for a given query, based on the number of longer views of the result compared with shorter views, passed to the ranking engine for future searches.
Why it matters for SEO: Describes learning from whether searchers stay on a result or quickly return. The takeaway is simple: answer the query well enough that people do not bounce back to the results. Practitioner practice
Which research papers explain link analysis and learning to rank?
The anatomy of a large-scale hypertextual Web search engine
What it describes: The prototype Google search engine: crawling, indexing, PageRank and the use of anchor text.
Why it matters for SEO: Google’s ranking systems guide links to this as “the original PageRank research paper”.[1] It also explains why anchor text describes the page it points to. Officially documented
Authoritative sources in a hyperlinked environment
What it describes: The HITS algorithm: hubs (pages that link to good sources) and authorities (pages linked to by good hubs) reinforce each other.
Why it matters for SEO: A useful mental model for topical authority: being linked by the recognised hubs of a topic.
When experts agree: using non-affiliated experts to rank popular topics
What it describes: The Hilltop algorithm: rank pages for popular topics by links from independent “expert” pages, ignoring links between affiliated sites.
Why it matters for SEO: Links from related sites you control count for little in this model. Independent, expert sources matter.
Topic-sensitive PageRank
What it describes: Computing several PageRank vectors biased toward different topics and combining them according to the topic of the query.
Why it matters for SEO: Relevance of the linking page’s topic can matter as much as its raw importance.
Learning to rank using gradient descent
What it describes: RankNet, a neural network trained on pairs of documents to learn a ranking function (from Microsoft Research).
Why it matters for SEO: Modern ranking combines many signals with machine learning, so no single factor is a lever on its own.
Accurately interpreting clickthrough data as implicit feedback
What it describes: An eye-tracking and click study showing clicks are biased by position and trust, but can be read as relative preferences between results.
Why it matters for SEO: Explains why raw CTR is a noisy signal and why titles and snippets matter. See on-page SEO.
Improving web search ranking by incorporating user behavior information
What it describes: Microsoft Research experiments showing that aggregated user behaviour features improved ranking accuracy.
Why it matters for SEO: Academic evidence that behaviour signals can improve ranking. It says nothing about how any production engine uses them.
What should you do with this?
- Earn links from important, topically relevant, independent pages. Placement in the main content matters more than footers.
- Use descriptive anchor text naturally. Do not buy links; Google’s spam policies cover link spam.[2]
- Satisfy the query so searchers stay. Keep key pages maintained.
Frequently asked questions
Is PageRank still used?
Google says PageRank was one of its core ranking systems at launch and that how it works has evolved a lot. Its ranking systems guide still lists link analysis systems and PageRank.
Has the PageRank patent expired?
Google Patents lists US 6,285,999 B1 as 'Expired - Lifetime'. Google Patents notes its legal status is an assumption, not a legal conclusion.
What is the reasonable surfer patent?
It is the informal name for US 7,716,225 B1, which describes weighting links using link features and user behaviour data, so that links more likely to be followed can count more.
See also
References
Pages accessed September 29, 2026 unless a date is given. See all sources and our editorial policy.
- ^ "A guide to Google Search ranking systems". Google Search Central.
- ^ "Spam policies for Google web search". Google Search Central. Updated May 2026 to cover generative AI responses.