The Google patents most useful to SEOs cluster into five groups: the foundational classics (PageRank US6285999B1, Reasonable Surfer US8117209B1, Site Quality Score US9031929B1, Information Gain US11354342B2, Historical Data US7346839B2), the AI-search and SERP patents behind AI Overviews, the user-behaviour and ranking-signal patents, the query-understanding patents, and the leaked systems (Navboost, Glue, siteAuthority) that have no public patent. Every patent below links straight to the live Google Patents page. A patent shows what Google has the right to do, not proof it is live in the algorithm, so read them as direction, not gospel.
I kept losing twenty minutes every time I wanted to re-find a specific Google patent. So I built the index I wished existed: the patents an SEO actually cites, grouped by what they do, each one a direct link to the real document. Bookmark this page.
The foundational classics (the ones everyone references)
Five patents underpin most SEO theory. If you only read five, read these.
| Patent | Number (links to patent) | What it covers |
|---|---|---|
| PageRank | US6285999B1 | The original ranking engine. Ranks pages by the link graph: a link is a vote, weighted by the linking page’s own authority. |
| Reasonable Surfer | US8117209B1 | Not all links pass equal weight. Links more likely to be clicked (position, anchor, font, context) pass more PageRank than buried footer links. |
| Information Retrieval Based on Historical Data | US7346839B2 | Freshness and the historical signals: document inception date, rate of content change, and anchor-text growth, used to fight stale pages and link spam. |
| Site Quality Score | US9031929B1 | The patent most tied to Panda. A site-wide quality score derived from the ratio of branded and navigational queries to a site against other signals. |
| Contextual Estimation of Link Information Gain | US11354342B2 | The information-gain signal. Scores how much NEW information a page adds beyond what the user has already seen. The anti-rehash patent. |
AI Overviews and modern SERP patents
The recent batch (2023 to 2024) that maps to AI Overviews, semantic retrieval, knowledge panels, and featured snippets. These are where the action is now.
| Patent | Number (links to patent) | What it covers |
|---|---|---|
| Generative summaries for search results | US11769017B1 | The AI Overview mechanic: how an LLM builds a generative summary on top of ranked search results. |
| Search results based on a compositional query | US11762933B2 | Matches complex multi-entity queries against the knowledge graph. Entity-relationship retrieval, not keyword matching. |
| Providing knowledge panels with search results | US11836177B2 | How entity knowledge panels get assembled from multiple sources. Relevant to entity SEO and being the cited authority. |
| Multi-source extraction and scoring of short query answers | US20230342411A1 | Scores candidate answers across multiple sources to produce a direct, featured-snippet-style answer. |
| Embedding based retrieval for image search | US11782998B2 | Neural embeddings for semantic image search. Visual relevance based on meaning, not just filename and alt text. |
| Evaluating an interpretation for a search query | US20230334045A1 | Validates how a query is interpreted using temporal and clustering features. The query-intent layer. |
User-behaviour and ranking-signal patents
Google has long been cagey about using clicks. These patents describe how behaviour data feeds ranking, and they line up with the Navboost system confirmed in the DOJ trial below.
| Patent | Number (links to patent) | What it covers |
|---|---|---|
| Modifying search result ranking based on implicit user feedback | US11816114B1 | The clicks-as-a-signal patent. Adjusts rankings using dwell time and engagement after the click. |
| Privacy-sensitive training of user interaction prediction models | US11741191B1 | Federated learning to predict user interaction without exposing individual user data. The modern, privacy-safe way to use behaviour signals. |
| Predictive query completion and predictive search results | US20230394072A1 | Predicts and pre-serves results before the query is finished. Shapes what demand even reaches your page. |
Query-understanding patents
How Google reads the query before it ranks anything. This layer maps directly to query fan-out in AI search.
| Patent | Number (links to patent) | What it covers |
|---|---|---|
| Combining parameters of multiple search queries | US11762848B2 | Combines parameters across related queries that share a line of inquiry. Maps to query fan-out in AI search. |
| Query categorization based on image results | US11782970B2 | Categorises a query using its image results and user behaviour. Cross-modal intent classification. |
| Search result filters from resource content | US11797626B2 | Generates SERP filters dynamically from the content of indexed resources. |
The leaked systems with no public patent
These are the names everyone asks for. They are not patents you can link to. They were confirmed through the May 2024 Content Warehouse API leak and the US Department of Justice antitrust trial testimony. Included because every SEO goes looking for them and comes up empty.
| System | Source (not a patent) | What it does |
|---|---|---|
| Navboost | DOJ trial + 2024 API leak | Re-ranks results using aggregated click data: good clicks, bad clicks, and the last-longest click, measured over a rolling window. |
| Glue | DOJ trial testimony | The same user-interaction signal applied across the whole SERP, including the non-web-result features. |
| siteAuthority | 2024 Content Warehouse API leak | An internal site-level authority score. Notable because Google publicly denied that a single "domain authority" style metric existed. |
How to read a Google patent without wasting your day
- A patent is permission, not proof. It shows what Google has the legal right to do. It is not confirmation the method is live, or live in the form described. Treat patents as directional.
- Read the Abstract and the Claims, skip the rest. The Abstract tells you the idea in a paragraph. The numbered Claims define what is actually protected. The 40 pages of "embodiments" in the middle are lawyer padding.
- Check the dates. A 2012 filing granted in 2024 tells you how long Google sat on an idea. Old filings can describe systems that shipped years ago or never shipped at all.
- Search by assignee yourself. On Google Patents, search
assignee:Googleplus your topic (for exampleassignee:Google information gain) to surface the family around any patent. - Cross-reference a human analyst. The late Bill Slawski (SEO by the Sea), Mike King (iPullRank), and Olaf Kopp do the translation work. Use their write-ups to sort the meaningful patents from the noise.
Frequently asked questions
Are these patents confirmation of how Google ranks?
No. A patent proves Google has the right to use a method, not that the method is live in the ranking system today. Some are implemented, some never shipped, some shipped in a different form. Read them as direction and intent, then validate with testing.
Which Google patent matters most for SEO right now?
For content, the Information Gain patent (US11354342B2): it rewards pages that add new information over pages that reword the consensus. For sites, the Site Quality Score patent (US9031929B1) and the implicit-user-feedback patent (US11816114B1), which line up with the leaked Navboost click system.
Is Navboost a patent I can read?
No. Navboost, Glue, and siteAuthority are systems confirmed through the 2024 Content Warehouse API leak and the US DOJ antitrust trial, not published patents. There is no Google Patents page to link. The closest public patents are the user-behaviour ones listed above.
How do I search Google Patents myself?
Go to patents.google.com and search assignee:Google plus your topic keyword, for example assignee:Google reasonable surfer. Sort by filing date, and read the Abstract and Claims first. The body text is mostly legal boilerplate.
What is the PageRank patent number?
PageRank is US6285999B1, titled "Method for node ranking in a linked database", filed by Lawrence Page at Stanford in 1998. It is the foundational link-based ranking patent and still the best starting point for understanding why links matter.
Soaring Above Search
Weekly AI search insights from the front line. One newsletter. Six sections. Everything that actually moved this week, with a practitioner's take.
Chief of Staff at StudioHawk, Australia's largest dedicated SEO agency. Specialising in AI search visibility, technical SEO, and organic growth strategy. Leading a team of 120+ across Melbourne, Sydney, London, and the US. Book a free consultation →