Lesson 1: Introduction to LLMs and What Drives AI Visibility

What Is GEO? How AI Search Works and Why It's Not SEO (Lesson 1) GEO (Generative Engine Optimization) is the practice of increasing a brand's visibility inside AI search tools like ChatGPT, Claude, and Perplexity. It is not the same as SEO. In Lesson 1, Nick Clark, Research Engineer at Gumshoe and PhD student at the University of Washington, explains why the two correlate weakly, how AI search systems actually retrieve and evaluate evidence, and what that means for brands trying to get cited. What is the difference between SEO and GEO? Traditional SEO optimizes for information retrieval: given a query, which pages are most relevant? AI search performs information seeking, which goes further and includes evaluation and synthesis of sources. Gumshoe's research team found only a weak correlation between SEO performance and GEO performance. High profile brands like Macy's, Samsung, and Motor Trend rank well in Google but underperform in AI search, while startups with thin content suites are cited frequently. Why does AI search recommend different brands than Google? Two reasons. First, personalization: models condition responses on conversation history and saved memories, so two people asking identical questions can receive entirely different brand recommendations. Second, stochasticity: the architecture of these systems creates a baseline level of noise. Published research on citation churn found AI search churns roughly 6x faster than Google search day to day. How do AI search systems actually work? Three components matter: Pre-training corpus. The web crawl used to train the base model. Models lean on parametric memory, so inclusion here creates an inductive bias toward your content. Retrieval (RAG / web search). Models issue subqueries to proprietary indexes and third party search engines to get past the knowledge cutoff date and surface long tail information. Memory and conversation history. These systems are not stateless. They accumulate context about each user and search over it to personalize recommendations. Is GEO just SEO fundamentals? Partly, since models do issue queries to search engines. But that claim is an incomplete picture. Most of the work happens at the evaluation and synthesis layer, where the criteria look far more human than backlink profiles and page metadata. Optimizing for AI search resembles public relations or analyst relations more than technical SEO, with the model acting as mediator. The sommelier analogy A wine list is retrieval. A sommelier asks what you're eating, your price range, and your preferred regions, then recommends. AI search is the sommelier. Key stat: McKinsey found AI search is the preferred method for consumers making product purchasing decisions, with baby boomers the only segment still preferring traditional Google search. They project $750 billion in purchases flowing through AI search by 2028. Chapters 00:00 What this lesson covers 00:17 McKinsey data on AI search and purchasing 01:00 Why SEO performance doesn't predict GEO performance 02:12 Information retrieval vs information seeking 04:59 The wine list and the sommelier 05:47 Why personalization changes everything 07:54 Variability, stochasticity, and the 6x churn rate 09:40 Pre-training corpus and parametric memory 10:46 Retrieval augmented generation and web search 13:54 Memory, conversation history, and statefulness 14:34 Google's ranking signals vs AI evaluation criteria (and published research on user research processes) 16:23 Is GEO really just SEO fundamentals? 17:16 Why GEO looks more like PR than SEO 18:02 TLDR for agencies Instructor Nick Clark, Research Engineer at Gumshoe and PhD student at the University of Washington. His first published paper examined how well models can identify the research process a given user would follow when making a purchase or booking travel, along with where those models fall short. That research directly informs how this course approaches personalization and AI visibility measurement. About this course A practical course on generative engine optimization from the research team at Gumshoe, where we measure AI visibility across models, personas, and market segments. Upcoming lessons cover personalization and persona level tracking, technical site optimization for AI crawlers, content formats models favor, and citation measurement. Subscribe for Lesson 2. #GEO #GenerativeEngineOptimization #AEO #AISearch #AnswerEngineOptimization #SEO #ChatGPT #AIVisibility #DigitalMarketing #LLM