Glossary · AI Search

Grounding

Also called: Source-Grounded Generation

Search-enabled assistants can retrieve current web sources before composing a response. Retrieval-Augmented Generation, or RAG, is one common grounding approach. The exact retrieval process, index, and citation behavior vary by product, and a source link does not mean the entire answer came from that page.

Grounding gives newer and smaller businesses a possible path into an answer because the system can consult current public pages instead of relying only on older training knowledge. It does not guarantee that a particular page will be retrieved or selected.

The practical requirement is eligibility: important pages should be crawlable by the relevant search systems, current, and clearly related to the question. After that, selection still depends on the engine, prompt, source set, and context.

A plain example

When Perplexity returns an answer with numbered links, those sources help ground that response. Opening the links shows what each source actually supports.