Wissenswertes

Key Takeaways: Segonzac analyzed a leak of Google Maps Geostore and found: 72 Geostore ranking signals 793 data source providers (along with mechanisms for provenance, priority, trust, and conflation = when several sources disagree) 446 local search intent types 50,998 Mapcore styles + 12,936 label styles = Mapcore decides what to show and what not 10,936 searchable Geostore declarations Google builds a canonical representation of the place that can incorporate data from mult

Key Takeaways: Metehan analyzed Common Crawl (CC) data and found a correlation to citation frequency in LLM Chats such as ChatGPT, Perplexity and others: Most major LLMs were trained on CC data (64% of models studied, 80%+ of GPT-3 tokens) CC prioritizes high-authority domains in its crawling via Harmonic Centrality These same domains tend to be cited most frequently by LLMs Feel free to use his tool to analyze Common Crawl authority scores for your domain or industry:...

Key Takeaways: Anderson analyzed the Google Leak & DOJ testimony: A deep dive mapping Google's Search Essentials and Helpful Content questions to leaked Content Warehouse attributes like Q*, siteAuthority, contentEffort, and predictedDefaultNsr. From Navboost to contentEffort he provides a clear understanding of how Google is or likely is working ICYMI: Back in 2024 the Website Boosting article from Mario Fischer visually outlined it pretty well Example I: EEAT decoded E-E-A-

Key Takeaways: Query Fan-Out auf Basis von Such-Ergebnissen ist sequentieller, sich wiederholender Prozess Es ist möglich diesen Prozess in den Request Responses eines Prompts nachzuvollziehen Sequentieller Prozess des Query Fan-Out in ChatGPT: Ausgehend vom ersten Prompt wird eine Suchanfrage erstellt Die Top Rankings zu dieser Suchanfrage werden analysiert und neue Sub-Suchanfragen angeleitet Jede Suchanfrage ist potenzieller Ausgangspunkt neuer Suchen und Ergebnisse Diese

Search Engines moved beyond keywords into understanding language in the vector space Maximum Inner Product Search (MIPS) aims at maximizing the inner product of a given query for a set of vectors