--- type: "Article" title: "information gain patent | topical authority content" description: "What Google's information gain patent (US11354342B2) actually describes, what it does not prove, and how to audit pages for redundancy." resource: "https://www.laurelinlabs.com/search-systems/information-gain-patent" tags: ["information-gain-patent", "topical-authority-content", "google-information-gain-score"] generated: { by: "laurelin-site-build/2026-09-02", at: "2026-09-02T12:00:00+00:00" } status: "stable" stale_after: "2027-03-02T00:00:00+00:00" --- # Information gain patent Google's patent 'Contextual estimation of link information gain' (US11354342B2, granted 2022) describes an information gain score: how much new information a document offers beyond documents the user has already viewed, predicted by a machine learning model. It is framed around an automated assistant and a user's own reading history. Google has not confirmed it is used in web ranking. For a site, the useful lesson is to make each page add something the reader has not already seen. ## What is the information gain patent? The information gain patent is [US11354342B2, Contextual estimation of link information gain](https://patents.google.com/patent/US11354342B2/en), assigned to Google with inventors Victor Carbune and Pedro Gonnet Anders. It has a priority date of October 18, 2018 and was granted on June 7, 2022, per Google Patents. Later patents with the same title, [US11720613B2](https://patents.google.com/patent/US11720613B2/en) and [US12013887B2](https://patents.google.com/patent/US12013887B2/en), are in the same family. ## How does the information gain score work? The patent describes a first set of documents the user has already seen and a second set of candidate documents. A machine learning model takes data from the documents and outputs, for each candidate, a score for how much new information it would add. The system then presents or ranks candidates by that score. The setting matters. Much of the description is about an automated assistant choosing what to read out or show next, based on what this user has already consumed. It is a per-user, per-session measure of novelty, not a fixed score for a page. ## What evidence is there that Google uses information gain? The evidence label is patent only. Google has not said the method runs in Search. The patent's claims and examples centre on an automated assistant, so reading it as proof of a web ranking factor goes beyond the text. What the SEO industry calls "information gain" (original content that adds to what already ranks) is a reasonable content principle, and Google's own [ranking systems guide](https://developers.google.com/search/docs/appearance/ranking-systems-guide) describes systems that reward original, helpful content. That is a separate claim from this patent being in production. ## What does the information gain patent mean for my site? Assume the reader has already seen the obvious answer, because in an assistant or AI search session they often have. A page that restates the same facts as the pages before it adds nothing measurable; a page with original data, a worked example or a clear decision rule adds something. This applies inside your own site too. If a reader moves through a series of your pages and each one repeats the previous page's introduction and advice, the later pages offer little gain over what that reader has already seen. A falsifiable test: take pages that a reader typically views in sequence (from analytics next-page paths) and measure the proportion of each page's sentences that also appear, or nearly appear, on the page before. High overlap means low gain for that reader. ## Which Laurelin audit checks test for information gain? The new check, [pages in a reading sequence repeat what the previous page said](https://www.laurelinlabs.com/audit/topical-content/374-pages-in-a-reading-sequence-repeat-what-the-previous-page-said), measures redundancy along real user paths, which is the setting the patent describes. It is a proxy: it does not reproduce Google's model. Related existing checks: - [Low information gain vs competitors](https://www.laurelinlabs.com/audit/topical-content/161-low-information-gain-vs-competitors): the cross-site version of the same idea. - [Commodity content](https://www.laurelinlabs.com/audit/topical-content/160-commodity-content): content that could have come from any source. - [Internal duplicate content](https://www.laurelinlabs.com/audit/topical-content/164-internal-duplicate-content): exact repetition across the site. - [Padded content](https://www.laurelinlabs.com/audit/topical-content/173-padded-content): length without added information. - [No first-hand experience or unique POV](https://www.laurelinlabs.com/audit/topical-content/162-no-first-hand-experience-unique-pov): first-hand detail is the easiest source of new information. ## What are the key dates for information gain patent? - 2018-10-18: Priority and filing date of US11354342B2 ([source](https://patents.google.com/patent/US11354342B2/en)) - 2022-06-07: US11354342B2 granted ([source](https://patents.google.com/patent/US11354342B2/en)) ## FAQ ### Is information gain a Google ranking factor? Not confirmed. Google holds a patent describing an information gain score, but it is framed around an automated assistant and a user's reading history, and Google has not said it is used in Search ranking. ### What does the information gain patent measure? How much additional information a document contains beyond documents the user has already viewed, estimated with a machine learning model. Hub: [Topical Content](/topics/topical-content.md)