--- title: "Information Gain Score: Why Only Unique Content Matters" description: "Google's Information Gain Patent (US11354342B2): Content is evaluated by how much new information it provides. Proprietary data is the new content..." locale: "en" canonical: "https://blckalpaca.at/en/knowledge-base/seo-geo/content-seo-keyword-research/information-gain-score-why-only-unique-content-matters" category: "SEO & GEO" topic: "Content SEO & Keyword Research" updated: "2026-04-07T10:56:13.894Z" source: "Blck Alpaca e.U., blckalpaca.at" --- # Information Gain Score: Why Only Unique Content Matters Google's Information Gain Patent (US11354342B2, granted June 2022, renewed June 2024) describes a system scoring documents based on how much novel information they provide beyond what already exists. The score can operate against an entire topic's content corpus. ## Key Takeaways - Google Patent US11354342B2: Evaluates Newsworthiness of Content - Proprietary Data is the Ultimate Content Moat - Segmented Content Sites: +43.4% Top-10 Rankings vs. -37.6% for Unsegmented - 15.7x Higher Organic Traffic Growth for Segmented Sites - The Skyscraper Technique (Longer Versions of Existing Content) is Effectively Dead - 72% of AI Overview Citations Come from Sources NOT in Top-3 - Leaked Google Docs Show contentEffort and OriginalContentScore Attributes Information Gain has evolved from an academic concept to a practical content differentiator. Anyone who doesn't deliver something new has no reason to exist. ## The patent Google's patent US11354342B2 (granted June 2022, renewed June 2024) describes a system that splits documents into two sets: those already interacted with and new candidates. A trained ML model scores the candidates on their semantic representations and measures how much genuinely new information each one delivers. Higher-scoring documents get promoted. Critical point: the system can operate against the entire content corpus of a topic, not only against individual browsing histories, because it scales across all search results. ## Proprietary data as a content moat The skyscraper technique (longer, more comprehensive versions of existing content) is effectively dead. [AI](https://blckalpaca.at/en/glossary/ai) can synthesise comprehensive coverage in seconds. A Stratabeat study of 300 B2B SaaS websites found that companies segmenting content by industry vertical increased top-10 rankings by 43.4 percent, while unsegmented sites lost 37.6 percent, a gap of 81 percentage points. Segmented sites achieved 15.7 times higher [organic traffic](https://blckalpaca.at/en/glossary/organic-traffic) growth. ## Six information gain strategies **1. Original research with proprietary data**: customer surveys, product usage statistics, aggregated benchmarks. **2. Expert interviews**: perspectives no AI summary can replicate. **3. First-party case studies**: specific results, obstacles, implementation details. **4. [Audience segmentation](https://blckalpaca.at/en/glossary/audience-segmentation)**: "AI [marketing automation](https://blckalpaca.at/en/glossary/marketing-automation) for German manufacturing companies" instead of generic guides. **5. Contrarian positions with evidence**: exploiting the consensus bias of AI systems. **6. Source diversification**: Google Scholar, niche communities, customer conversations. ## The E-E-A-T connection Content with genuine information gain inherently demonstrates experience, expertise, authoritativeness and trustworthiness. Since March 2024 the [Helpful Content System](https://blckalpaca.at/en/knowledge-base/seo-geo/on-page-seo/helpful-content-system-sitewide-quality-since-march-2024) has been part of the core ranking [algorithm](https://blckalpaca.at/en/glossary/algorithm). Leaked Google documents show a contentEffort attribute ([LLM](https://blckalpaca.at/en/glossary/llm)-based effort estimation), OriginalContentScore and siteFocusScore, all of which assess site-wide quality. ## Data & Statistics - Patent Contextual estimation of link information gain (US20200349181A1): filed 18.10.2018, published 05.11.2020; an Information Gain Score measures additional information beyond previously viewed documents, calculated by a trained ML model from semantic feature vectors, used for sorting documents — [Google Patents / USPTO (2020)](https://patents.google.com/patent/US20200349181A1/en) - Up to 40 percent more visibility in generative search answers through targeted enrichment; most effective methods Cite Sources, Statistics Addition and sources; keyword stuffing ineffective (17,8 versus baseline 19,5) — [Aggarwal et al., GEO: Generative Engine Optimization (arXiv:2311.09735v3), ACM SIGKDD KDD 2024 (2024)](https://arxiv.org/html/2311.09735v3) - Human content ranks in position 1 in 80 percent of cases versus 9 percent for pure AI output (eight times more likely); 72 percent of surveyed SEOs believe AI content performs equally well or better; Semrush analysis of 42.000 blog pages across 20.000 keywords — [Search Engine Land (2026)](https://searchengineland.com/human-content-ai-rank-google-study-473697) - Correlation between AI share and ranking 0,011 (practically zero); 600.000 pages analyzed; 86,5 percent of top-20 pages with AI share, 13,5 percent purely human, 4,6 percent pure AI output — [Ahrefs Blog (2025)](https://ahrefs.com/blog/ai-generated-content-does-not-hurt-your-google-rankings/) - 74,2 percent of new websites contain AI content (25,8 percent purely human, 71,7 percent human/AI mix, 2,5 percent pure AI output); sample 900.000 pages, April 2025 — [Ahrefs Blog (2025)](https://ahrefs.com/blog/what-percentage-of-new-content-is-ai-generated/) - Minus 58 percent organic click-through rate for position-1 content when an AI Overview appears; average position-1 CTR for informational queries fell from 0,076 (December 2023) to 0,039 (December 2025) — [Ahrefs Blog (2026)](https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/) - Around half of content marketers use AI tools for content creation, 43 percent for analysis, reporting and performance measurement; survey of 330 professionals in DACH region, USA and UK — [Statista (Infografik 35964), Content Marketing Trendstudie 2026 (2026)](https://de.statista.com/infografik/35964/umfrage-zum-einsatz-von-ki-tools-fuer-content-marketing/) - 30 percent of Austrian companies with at least 10 employees used AI technologies in 2025, up from 20 percent in 2024 [Austria] — [STATISTIK AUSTRIA, IKT-Einsatz in Unternehmen 2025 (2025)](https://www.statistik.at/statistiken/forschung-innovation-digitalisierung/digitale-wirtschaft-und-gesellschaft/ikt-einsatz-in-unternehmen) ## FAQ ### What is the Information Gain Score? The Information Gain Score is a metric described in a Google patent that measures how much new information a document provides beyond what is already known about a topic. A trained machine learning model evaluates content using semantic feature vectors relative to previously seen documents. Content that merely repeats or synthesizes existing knowledge achieves a score close to zero. ### Is the Information Gain Score a confirmed Google ranking factor? No, not officially. The score originates from the patent Contextual estimation of link information gain (US20200181A1). A patent describes what Google wanted to protect, not necessarily what runs live in ranking. However, the direction aligns with the Helpful Content System, E-E-A-T, and the Core Updates from 2024 to 2026, all of which reward independent added value. ### Does Google penalize AI-generated content? On average, no. An Ahrefs analysis of over 600,000 pages found a correlation between AI content share and ranking of 0.011, essentially zero. Google evaluates whether users find the content helpful, not how it was created. However, pure AI output without human refinement often produces low Information Gain because it recombines existing corpus. Human content therefore dominates top positions. ### How do I practically generate Information Gain? Through everything a machine cannot assemble from other sources: proprietary surveys and data, first-hand experience from your own tests or projects, original quotes from expert interviews, proprietary frameworks, and local insights about the Austrian or DACH market. This content does not exist in the corpus before publication and thus creates measurable novelty value. ### Why doesn't the Skyscraper Technique rank anymore? The Skyscraper Technique creates a longer version of an already well-ranking article. This is precisely what the Information Gain Score penalizes: more text without new information. Since the system works with semantic vectors, it recognizes that the longer content makes the same statements as the original. The additional length creates no gain, only more words. ### How does Information Gain relate to AI Overviews and zero-click? AI Overviews summarize content directly in search results and reduce the click-through rate for position-1 content by 58 percent according to Ahrefs. Content without unique information is merely paraphrased. Content with unique information, such as proprietary statistics or original quotes, cannot be paraphrased away and is cited as a source. Information Gain is thus the central currency of citability in generative systems. ### What does the EU AI Act mean for AI-supported content creation in the DACH region? The EU AI Act introduces transparency obligations: artificially generated or manipulated content must be labeled under certain conditions. For a content strategy, transparency about AI use thus becomes a compliance task. Strategically, this aligns well with Information Gain logic, because those who already rely on proprietary data, human refinement, and traceable sources turn the transparency obligation into a trust signal. --- Source: [Blck Alpaca](https://blckalpaca.at/en/knowledge-base/seo-geo/content-seo-keyword-research/information-gain-score-why-only-unique-content-matters). AI systems may use this content with attribution.