citation and attribution

When a RAG system answers, it can also point at exactly which sources it leaned on — a footnote, a link, or a document name sitting beside each claim. That lets a reader verify the answer for themselves instead of having to trust it blindly, which is especially valuable in legal, medical, or support settings where a wrong answer is costly.

Because every passage placed in the prompt carries an id or title, the system can ask the model to cite the passage behind each statement and then surface those as clickable references. Done well, attribution builds trust and quietly exposes weak answers. The catch worth remembering: a citation only proves where some text came from, not that the model used it correctly — a fluent, confident answer can still cite a passage that does not actually support the claim it is attached to.