Chapter 2 of 11 · Part 1: The Shift
From Rankings to Recommendations: How LLMs Choose What to Mention
AI answers are produced in two phases: retrieval (which sources make the candidate set) and generation (which passages are woven into the answer). Both phases must work in your favour.
What does chapter 2 cover?
AI answers are produced in two phases: retrieval (which sources make the candidate set) and generation (which passages are woven into the answer). Both phases must work in your favour.
The book closes every chapter with a summary box. This is chapter 2's, in full and in the book's own words.
- AI answers are produced in two phases: retrieval (which sources make the candidate set) and generation (which passages are woven into the answer). Both phases must work in your favour.
- Ranking number one on a traditional results page does not guarantee inclusion in an AI-generated answer. The selection criteria are different and query-dependent.
- AI systems draw on three knowledge sources: training data (fixed), real-time retrieval (live), and grounding data (structured). A strong brand is present in all three.
- The same query produces different answers across Google AI Overviews, ChatGPT, and Perplexity, but the brands that appear across all three share common attributes: entity resolution, corroboration, and extractable content.
- ECHO's four pillars (Entity, Corroboration, Hooks, Output) map directly to the prerequisites that AI answer engines impose for inclusion.
How does chapter 2 open?
The most common misunderstanding about AI-generated answers is that the AI "just knows things." It does not. Or rather, it knows things the way a person who read a million documents once might know things: hazily, with a strong tendency to fill gaps with plausible-sounding guesses. The architecture that matters for this book is the one that corrects for that weakness, and it has two distinct phases that are worth understanding separately.
The opening paragraph of the chapter, quoted from the book.
What is in chapter 2?
Chapter 2 runs to 5 sections, starting on page 15.
- 1Retrieval vs. Generation: What Is Actually Happening Under the Hood
- 2Why Ranking Number One No Longer Guarantees Being Mentioned
- 3The Role of Training Data, Real-Time Retrieval, and Grounding Sources
- 4Case Example: Same Query, Three Different AI Answers
- 5What This Means for the Rest of This Book
Where does chapter 2 sit in ECHO?
Chapter 2 sits in Part 1, which establishes what has changed before the four pillars are introduced. It belongs to no single pillar.
- Chapter
- 2 of 11
- Part
- 1: The Shift
- Pillar
- None, this chapter spans the framework
- Starts on page
- 15
- Sections
- 5