Retrieval-augmented generation
An architecture where a system retrieves relevant documents or passages first and then has a language model generate an answer conditioned…
GEO AISearch grounding
Grounding
Constraining a language model's output to information retrieved from a specified source — a search index, a document set or a database — so answers can be attributed and verified.
In full
Grounding is what separates a search-connected assistant from a model answering purely from parametric memory: retrieved passages are placed in the model's context and the answer is generated over them, usually with citations back to the source documents. It substantially reduces but does not eliminate hallucination, since a model can still misread or over-generalise retrieved text. For site owners, being retrievable is the prerequisite for being grounded on, which is why crawlability and clear factual statements matter.
Example
An assistant asked about a tax threshold retrieves the current guidance page and quotes the figure with a link rather than recalling last year's number.
Related terms
Hallucination
A confident, fluent but factually wrong or fabricated statement produced by a language model, including invented citations, prices…
Chunk retrieval
Retrieving individual segments of a document rather than whole pages, which is how most AI search and RAG systems select what to feed a…
AI Overviews
Google's AI-generated summaries shown at the top of many search results, synthesising information from multiple web pages with links to…
Citation share
The proportion of AI answer citations across a tracked prompt set that point to a given domain, used as the AI-era analogue of ranking…
Where to read more
- Google Crawler (User Agent) Overview · Documentation · developers.google.com