Retrieval-augmented generation (RAG)

Also: RAG

RAG (retrieval-augmented generation) is the technique where an AI model fetches relevant external documents at query time and grounds its answer in them, instead of relying only on training data. It is why fresh, crawlable web content can be cited by AI search without being in a model's training set.

Reviewed June 2026

Why it matters

RAG is the reason publishing matters again: a distributor that puts clean spec and cross-reference data on crawlable pages can be retrieved and cited today, without waiting to be absorbed into a future model training run.

In practice

W.W. Grainger, the industrial/MRO distributor, built a RAG-based search system on Databricks Mosaic AI and Vector Search over its catalog of roughly 2.5 million products with about 400,000 daily updates. At query time the assistant retrieves the relevant, freshly synced product records and grounds its answer in them, giving sales and call-center agents conversational product discovery instead of answers a static model might hallucinate from stale training data.

Source: databricks.com

Related terms

15 minutes. The one constraint.

Book a 15-minute strategy call. No sales pitch — we’ll name the single constraint capping your growth right now and the one change with the highest payback.

Book a strategy call Or get the free written audit

Retrieval-augmented generation (RAG)

Why it matters

In practice

Grounding

LLM citation

Generative engine optimization (GEO)

15 minutes. The one constraint.