What is passage retrieval?
Passage retrieval is the step where a search or AI system selects a specific relevant passage from an indexed document to answer a query, rather than returning the whole document.
Chunking and passage retrieval work together but happen at different times. Chunking is preparation: the document gets split up in advance. Passage retrieval is the selection: at the moment someone asks a question, the system picks which of those pre-made pieces actually answers it.
This is why a page can rank fine and get crawled without any issue, yet still never get cited anywhere. If none of its individual passages read as a self-contained answer, there's nothing for the retrieval step to select, even though the page as a whole is clearly relevant to the topic.
The misunderstanding is assuming the whole page gets judged each time a citation happens or doesn't. Usually the unit being judged is a single passage. A practical test is whether a given paragraph would still make sense pasted into a conversation with no other context attached, if not, it probably won't survive retrieval intact either.
Related
- ChunkingChunking is the process of splitting a document into smaller pieces before it's indexed, so a search or AI system can retrieve the one relevant piece instead of the whole page.
- GroundingGrounding is when a model answers using documents retrieved at the time of the question, rather than only from what it absorbed during training.
- Retrieval-augmented generation (RAG)Retrieval-augmented generation is the technique of retrieving relevant documents and giving them to a model as context before it answers.
Want to know where you actually stand on this? Run a free visibility check or try the free tools.