3.5Intermediate8 min

Keywords: From Strings to Entities and Semantic Understanding

Lucas Blochberger·7 April 2026·Updated 11 June 2026

Definition

Keyword optimization has shifted from density to semantic understanding. BM25 remains the basis of Google's retrieval pipeline, transformer embeddings supplement for semantic re-ranking. Content recognized as an entity in the Knowledge Graph appears 50 percent more likely in Rich Results. Entity salience — how central an entity is to a text — directly influences rankings.

Key Takeaways

✓Two stages: BM25 for fast retrieval, transformers for semantic re-ranking
✓Entity content: 50% more likely in Rich Results
✓Entity salience: Position in text, frequency, grammatical dependency
✓Focus each URL on one dominant entity
✓92% of SEO professionals: Search intent is critical for ranking success
✓Only 5% of web content is in German — less competition
✓Only 5% of web content is in German — less competition

Keyword optimization has fundamentally changed — from string matching to semantic understanding.

BM25 as Foundation

BM25 remains the foundation of Google's retrieval pipeline. Modern pipeline: BM25 for fast initial retrieval, then Transformer embeddings for semantic re-ranking. Tools like SISTRIX WDF*IDF remain useful for benchmarking but should guide rather than prescribe.

Entity-Based SEO

Content recognized as an entity in the Knowledge Graph is 50 percent more likely to appear in Rich Results. Teams should map persons, products, brands, and core concepts to Wikidata Q-IDs or Knowledge Graph entries.

DACH Keyword Specifics

German compound words: Google handles splitting effectively. Umlauts correct in content (ä, ö, ü), vowel+e in URLs (ae, oe, ue). Never simply drop the dots: "Über" and "Uber" are different words. Formal "Sie" for B2B credibility. Only 5 percent of web content is in German — less competition.

Data & Statistics

Knowledge Graph zum Start: mehr als 500 Millionen Objekte und mehr als 3,5 Milliarden Fakten und Beziehungen

Google The Keyword (offizieller Google-Blog), Introducing the Knowledge Graph: things, not strings (2012)

Etwa 15 Prozent aller Suchanfragen hatte Google noch nie zuvor gesehen; RankBrain 2015 eingefuehrt

Search Engine Journal, Google Algorithm History: RankBrain (2024)

BM25 Standardparameter k1 = 1,2 und b = 0,75 (Termfrequenz-Saettigung und Dokumentlaengen-Normalisierung)

Elastic Blog, Practical BM25 - Part 3: Considerations for Picking b and k1 in Elasticsearch (2018)

Salience-Score im Bereich 0 bis 1,0; Werte nahe 1,0 = hohe Zentralitaet der Entitaet fuer das Dokument

Google Cloud Natural Language API Documentation (Entity) (2024)

JSON-LD auf 41 Prozent der mobilen Seiten (2022: 34 Prozent); WebSite 12,73 Prozent, Organization 7,16 Prozent, LocalBusiness 3,97 Prozent

Web Almanac 2024 - Structured Data (HTTP Archive) (2024)

Featured Snippet ueber Position 1 erhaelt ca. 8,6 Prozent der Klicks vs. ca. 19,6 Prozent fuer das Ergebnis darunter; ca. 12,29 Prozent der Anfragen zeigen ein Featured Snippet (Analyse von 2 Mio. Snippets)

Ahrefs, Study of 2 Million Featured Snippets (2017)

GEO steigert Sichtbarkeit in generativen Antworten um bis zu 40 Prozent; Zitate +41 Prozent, Statistiken +33 Prozent, Quellenbelege +28 Prozent relativ (10.000 Anfragen, GEO-Bench)

arXiv, Aggarwal et al., GEO: Generative Engine Optimization (2024)

AI Overviews: Hoechstwert knapp unter 25 Prozent der Anfragen (Juli 2025), Rueckgang auf unter 16 Prozent (November 2025); Basis mehr als 10 Mio. Keywords (Semrush)

Search Engine Land, Google AI Overviews surge and pullback (2025)

E-E-A-T = Experience, Expertise, Authoritativeness, Trustworthiness; Trust ist die wichtigste Komponente

Google Search Central Blog, E-A-T gets an extra E for Experience (2022)

Durchschnittliche Wortzahl eines Google-Top-10-Ergebnisses: 1.447 Woerter (Analyse von 11,8 Mio. Suchergebnissen); keyword-ueberladene Title-/H1-Tags ohne Ranking-Korrelation

Backlinko, We Analyzed 11.8 Million Google Search Results (2025)

Suchmaschinen-Marktanteil Oesterreich: Google 81,87 Prozent, Bing 9,01 Prozent, DuckDuckGo 2,75 Prozent (Mai 2026)

StatCounter Global Stats - Search Engine Market Share Austria (2026)

8,69 Mio. Internetnutzer in Oesterreich, Online-Penetration 95,3 Prozent (Januar 2025)

DataReportal, Digital 2025: Austria (2025)

FAQ

What is Semantic SEO?

Semantic SEO is the optimization of content for meaning and entities rather than keyword density. Search engines process terms via the Knowledge Graph as linked concepts—things, not strings. The goal is to cover a topic comprehensively and be recognized as a distinct entity, not to place a keyword as often as possible.

Is keyword density still a ranking factor?

No, keyword density as a lever is outdated. Google's retrieval is based on BM25, which rewards additional repetitions of a term ever more weakly via term-frequency saturation and balances longer texts through document-length normalization (default values k1 = 1.2 and b = 0.75). A Backlinko analysis also found no correlation between keyword-stuffed title or H1 tags and better rankings.

What is an entity in the SEO context?

An entity is a uniquely identifiable concept such as a company, product, person, or technical term. In the Knowledge Graph, each entity receives an ID and is linked with attributes and other entities. This allows Google to disambiguate ambiguous terms, such as Jaguar as an animal, car brand, or operating system.

What is entity salience and how do I increase it?

Entity salience describes how central an entity is to a text. Google's Natural Language API outputs a score between 0 and 1.0, where values close to 1.0 indicate high centrality. Salience can be increased by early placement of the main entity in title, H1, and introduction, through deliberate repetition in moderation, and by mentioning it as the subject or object of key statements.

How do I anchor an entity in the Knowledge Graph?

Through structured data according to Schema.org, a page is informed about which entity it concerns. The sameAs attribute links to authoritative references such as Wikipedia, Wikidata, or official profiles. Consistent spelling and linking across multiple sources additionally stabilize the entity. JSON-LD is now standard on the web and present on 41 percent of mobile pages.

How do I optimize content for AI Overviews and generative answers?

The field is called Generative Engine Optimization (GEO). A study across 10,000 queries shows a visibility increase of up to 40 percent. The most effective levers are adding citations (approx. 41 percent), statistics (approx. 33 percent), and source references (approx. 28 percent) relative improvement. In practice, this means: link trustworthy sources, support statements with numbers, and deliver precise, extractable answers.

What role does E-E-A-T play in Semantic SEO?

E-E-A-T stands for Experience, Expertise, Authoritativeness, and Trustworthiness. Google calls trust the most important component. On-page, this means marking authors as their own entity via Author schema, citing sources, and setting trust signals such as update date, imprint, and technical references. An author entity consistently linked with topic and brand strengthens the credibility of the entire page.

How does your website perform?

Get a free, AI-powered SEO report of your website by email – technical SEO, on-page, keywords & competitors. No obligation.

Get a free SEO audit →

Previous← E-E-A-T On-Page: 85% of AI Citations Show 3+ E-E-A-T Signals NextInternal Linking: 23 Billion Links Show 4x Traffic Effect →