Chunking is a key step in the RAG pipeline; common splitting methods include splitting by fixed length, by paragraph boundaries, and by semantic similarity, and chunk size requires trade-offs.