Passage Embeddings

Passage Embeddings are embedding vectors created for individual fragments (passages) of a page, enabling Google to index specific sections rather than entire documents. Google creates separate passage embeddings for H2 sections, allowing each to rank independently.

For example, a 'Water Slides' section in an aquapark article has its own embedding and ranks for slide-related queries. This mechanism powers both the Passage Ranking system by matching query vectors to passage vectors, and AI Overview's fragment selection by identifying the most relevant passage embeddings for citation.

This understanding leads to specific optimization strategies: each H2 section should be designed as an autonomous unit of knowledge with BLUF at the beginning. This means including the entity's full name (not pronouns), a concrete conclusion upfront, and Atomic Claims in the body. The goal is to create sections that function independently of other page content.