SEO patents and research papers
SEO patents and research papers are the primary documents behind how search engines are thought to work. Patents describe methods a company has legally protected; research papers describe methods scientists built and tested. This section lists 41 patent documents and 42 papers, grouped into seven themes, each checked against Google Patents, arXiv, the ACL Anthology or the publisher’s DOI record. A patent never proves a method is in use.
What are SEO patents and research papers?
A patent is a legal right: a U.S. patent gives its owner the right to “exclude others from making, using, offering for sale, or selling” an invention.[1] Search companies patent many ideas they never ship, and patents describe methods in deliberately broad terms. A research paper reports a method and experiments, usually in a peer-reviewed venue or as an arXiv preprint. Papers by Google researchers show what Google has studied, not what runs in Search.
Why do patents and papers matter for entity SEO?
They show the problems search engines work on (disambiguating entities, judging quality, matching meaning, grounding AI answers) and the kinds of signals that could solve them. Used carefully, they help you form hypotheses and explain why good practice works: consistent entity facts, earned links, clear structure. Used carelessly, they spread myths. Practitioner practice
Browse by theme
| Theme | Covers | Patents | Papers |
|---|---|---|---|
| Entities and Knowledge Graph | Fact repositories, entity types, disambiguation, salience, Freebase, Wikidata | 8 | 7 |
| Ranking and link analysis | PageRank, reasonable surfer, seed distance, historical data, clicks, HITS, Hilltop | 9 | 7 |
| Query understanding and semantic matching | Synonyms, concepts, heading-text answers, phrase indexing, word2vec, DSSM, BM25 | 5 | 5 |
| Quality, trust and E-E-A-T | Panda, site quality score, Agent rank, TrustRank, knowledge-based trust | 4 | 3 |
| Local search | Location prominence, distance demotion, map spam, local item extraction | 7 | 0 |
| Neural IR, BERT, MUM and passages | Transformer, BERT, T5, answer passages, heading vectors, dense retrieval | 3 | 10 |
| AI answers and RAG | Generative summaries, information gain, REALM, RAG, citations, GEO | 5 | 10 |
| Total | 41 | 42 | |
How should SEOs read a patent?
- Check who owns it and its status. Google Patents shows the assignee and a legal status (for example “Active” or “Expired”), which it calls an assumption, not a legal conclusion.
- Read the abstract, then the claims. The claims define what is protected; the description adds examples that may never have been built. Practitioner practice
- Separate the dates. The priority date is the earliest filing the patent relies on, the filing date is when this application was filed, and the grant date is when it issued. Ideas are often years older than the grant.
- Look for the family. Later patents with the same priority date (continuations) often repeat the same abstract with new claims.
- Know an application from a grant. A published application (number ending in A1) has not been granted and may never be.
- Look for confirmation. Only Google’s own statements, such as its ranking systems guide, confirm that a system exists.[2]
- Treat it as a hypothesis. Test ideas on your own sites before acting on them.
Which Google systems are confirmed, and which documents relate to them?
| Confirmed system (Google) | Related documents on this wiki | Link to Google’s confirmation |
|---|---|---|
| PageRank / link analysis | US 6,285,999 B1; Brin and Page (1998) | Google links both from its guide Officially documented |
| BERT | Devlin et al. (2019) | Launched in Search in 2019[3] |
| MUM | T5 (Raffel et al., 2020) | MUM “uses the T5 text-to-text framework”[4] |
| Passage ranking | Answer-passage patents (related, not confirmed as the system) | Listed in the ranking systems guide |
| RankBrain, neural matching | No patent or paper identified by Google; see query understanding | Listed in the ranking systems guide |
| Knowledge Graph | Entity patents and papers | Launched May 2012[5] |
Which documents are most useful for entity SEO?
- Finding and disambiguating references to entities on web pages: why related entities disambiguate you.
- A New Entity Salience Task: making your main entity central.
- Anchor text summarization for corroboration: links that name your entity.
- Scoring local search results based on location prominence: citations, mentions and reviews.
- Context scoring adjustments for answer passages: why heading structure matters.
- Generative summaries for search results: why AI answers start with retrieval. See also from SEO to AEO and AEO.wiki’s research page.
How was this list verified?
Every patent’s number, title, assignee, inventors, priority, filing and grant or publication dates and legal status were taken from its Google Patents page on September 29, 2026. Every paper’s title, authors, venue and year were checked against arXiv, the ACL Anthology, the Crossref record for its DOI or the official proceedings page. Summaries are written from each document’s abstract and description. Anything we could not verify was left out:
- “The PageRank Citation Ranking: Bringing Order to the Web” (Stanford technical report): the Stanford InfoLab server returned errors, so we could not confirm its record. Google instead links the 1998 Brin and Page paper as the original PageRank paper, which is listed.
- RankBrain, neural matching and other named systems: we found no patent or paper that Google identifies as describing them, so none is attributed.
Assignee names are shown as Google Patents lists them today (for example, Google LLC for patents originally filed by Google Inc.). Report errors through the contact page; see our editorial policy.
Frequently asked questions
Do Google patents prove how Google ranks pages?
No. A patent shows that a method was legally protected, not that it is used, used as written or still used. Only Google's own statements confirm specific systems.
How many patents and papers does EntitySEO.wiki list?
41 patent documents (40 granted patents and 1 published application) and 42 research papers, across seven themes.
How do I look up a patent?
Search the number on Google Patents (patents.google.com), for example US6285999B1, and check the assignee, dates, legal status, abstract and claims.
What is the difference between a patent application and a granted patent?
An application (US numbers ending in A1) has been published but not granted and may never be. A granted patent (ending in B1 or B2) has issued.
Are Google research papers used in Search?
Only when Google says so. Google has confirmed BERT in Search, and says MUM uses the T5 framework. Most other papers show research directions.
See also
References
Pages accessed September 29, 2026 unless a date is given. See all sources and our editorial policy.
- ^ "Patent essentials". USPTO.
- ^ "A guide to Google Search ranking systems". Google Search Central.
- ^ "Understanding searches better than ever before (Pandu Nayak)". Google (The Keyword). Published October 25, 2019.
- ^ "MUM: A new AI milestone for understanding information (Pandu Nayak)". Google (The Keyword). Published May 18, 2021.
- ^ "Introducing the Knowledge Graph: things, not strings (Amit Singhal)". Google (The Keyword). Published May 16, 2012.