Deployment & Scaling · Practitioner
Grounding and RAG: Agent Search, RAG Engine and Grounding with Google Search
Four different ways to stop a model inventing facts, from a one-line grounding flag to a fully custom retrieval pipeline — and how to choose between them on control, effort and citation quality.
- 1.4 hours lesson
- intermediate level
- Google Cloud AI Track
- Narrated presentation
What you’ll learn
- Four grounding rungs: Google Search grounding, Agent Search, RAG Engine, fully custom
- Grounding with Google Search returns grounding metadata you are required to display
- Agent Search as the managed default for enterprise document corpora
- Retrieval failure vs generation failure — measured separately with recall@k and faithfulness
- Citations and source URIs as a product requirement, not a nicety
- Structure-aware chunking with metadata for filtering and precise citation
Inside this lesson
A narrated, slide-by-slide presentation with 1 prediction exercise, 1 knowledge check, 4 videos. You answer as you go, and the platform checks your work.