HeyJaredSEO Research Suite - The thought leading podcast for Generative Engine Optimization and SEO
1-2 times a week in the podcast are discussed Google patents, research papers and other hot topics like E-E-A-T, LLMO, Generative Engine Optimization (GEO), semantic search and Ranking.
- Indexed episodes, last 90 days
- 5
- Latest publication
- Sep 17, 2026
- Audience
- Checking…
- Earliest in this view
- Jul 7, 2026
Latest episodes
GEO Patent of the week: Generating and Utilizing compressed grounding data data for search engines that utilize generative artificial intelligence models (opens the original)
Read excerpt
This patent application details the Grounding Compression System , a framework designed to optimize how search engines feed web data into generative AI models. To overcome token limits and information overload, the system intelligently prunes retrieved content using a multi-step pipeline focused on freshness, credibility, and semantic relevance . It employs advanced techniques like extractive summarization and dense vector embeddings to distill vast amounts of information into a compact format t
GEO patent of the week: Determining Generative Search Resault Document for queries using Generative Artificial Intelligens Models and Arbitration Models (opens the original)
Read excerpt
This Microsoft patent describes a generative document system designed to evolve traditional search results into dynamic, AI-produced answers. Instead of a simple list of links, the technology creates multiple candidate documents simultaneously, including comprehensive summaries , visual digests , and direct answers . A specialized arbitration model then evaluates these versions based on accuracy, utility, and visual appeal to present the highest-quality result to the user. To maintain efficiency
GEO patent of the week: Knowledge Graph Query Optimization for Retrieval Augmented Generation (opens the original)
Read excerpt
This Microsoft patent introduces an advanced two-stage Retrieval-Augmented Generation (RAG) system designed to improve how AI chatbots answer questions. By moving away from traditional keyword-based searches, the framework uses a structured knowledge graph to minimize irrelevant data and maximize factual precision. The process utilizes two distinct large language models : one to translate conversational intent into a technical graph query and another to generate a natural, grounded response. Thi
Search patent of the week: Generative Search Results Documents Based on Enhanced Search Results Using Generative AI Models (opens the original)
Read excerpt
This Microsoft patent outlines a generative document system designed to replace traditional link lists with structured, AI-curated reports. To ensure depth, the technology uses a query fan-out method to expand a single user request into multiple related sub-queries. A specialized architecture then employs multiple small AI models to rank results, generate concise summaries, and assemble topic-specific sections in under ten seconds. The system maintains accuracy through quality control gates that
Search patent of the week: Utilizing large language model (LLM) in responding to multifaceted queries (opens the original)
Read excerpt
This episode examines a Google patent designed to improve how search engines process complex, multifaceted, or noisy natural language queries . The system uses a Large Language Model (LLM) to "fan out" a single complicated request into several distinct subqueries that target specific facets of the user's intent. To ensure efficiency and accuracy, these subqueries are filtered using relatedness and diversity metrics before the system retrieves and synthesizes the final search results. This techno
Publishing over time
Last 90 days. Choose a month to open its work.
Recurring subjects
Named in the text we hold. One piece can cover several.
Audience
No verified audience measurement yet.
About this data
Counts cover the work we have indexed. Tone needs enough text and a confident classification. Excerpts and episode notes are not full articles or transcripts.
Identity or attribution wrong? Suggest a correction.