Source Attribution
Also called Provenance
Tracking and showing where each piece of an answer came from.
Think of it like
A recipe card noting which cookbook each step was adapted from.
Example
An answer lists "Sources: Billing Policy §3, Onboarding Guide p.7" so a user knows exactly what informed it.
How it actually works
Attribution means every retrieved chunk carries provenance — document, section, URL, timestamp — through the pipeline so the final answer can name its sources. It supports citations, auditing, and access control, and it is only as good as the metadata captured at ingestion. Dropped provenance early on cannot be recovered later, which is why it is a design concern from the start.
For product teams
Essential for regulated or high-trust domains where "says who?" must have an answer.
For engineers
Propagate document/section/URL metadata end-to-end; provenance lost at ingestion is unrecoverable downstream.
Related
- Citation Grounding — The mechanism behind citation grounding.
- Ingestion — Depends on metadata kept through ingestion.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome