Skip to content
GigAI Tools
Definition

What is Retrieval-Augmented Generation (RAG)?

Also known as Retrieval-Augmented Generation (RAG).

Definition

Retrieval-Augmented Generation (RAG) is how an AI answers using sources it was never trained on: it splits documents into chunks, finds the chunks that match the question, and hands only those fragments to the model. It is why your page is read as pieces rather than as a whole.

  • Documents are split before they are stored, typically by heading section and then into token windows of a few hundred tokens with a small overlap.
  • The model sees one chunk at a time, without your heading hierarchy, your introduction or the paragraph above it. A sentence starting 'It also supports…' becomes meaningless.
  • This is why self-contained writing matters more for AI than for humans. Naming the subject again instead of writing 'it', and deleting phrases like 'as mentioned above', directly changes whether a fragment is usable.
  • Sections longer than the chunk window get split, and only the first piece keeps its heading. Shorter sections are therefore safer.

Frequently asked questions

Try Retrieval Chunk Preview

Free, private and instant. Everything runs in your browser.