{
  "@context": "https://schema.org",
  "@type": "DefinedTerm",
  "schema_version": "1.1",
  "content_item_id": "dictionary-rag.en",
  "translation_group_id": "dictionary-rag",
  "locale": "en",
  "type": "glossary",
  "section": "dictionary",
  "slug": "rag",
  "title": "RAG",
  "description": "What retrieval-augmented generation does and where the term stops.",
  "direct_answer": "RAG, or retrieval-augmented generation, is an architecture in which a system first retrieves relevant context from an external corpus and then supplies it to a model for answer generation. RAG does not automatically imply the open web, fresh data, factual accuracy, or visible citations.",
  "sections": [
    {
      "heading": "Working meaning",
      "paragraphs": [
        "RAG joins two stages: retrieval or selection of passages from a defined corpus, followed by answer generation conditioned on those passages. The corpus may contain company documents, a knowledge base, or public web pages; its boundary comes from the implementation, not from the term itself."
      ],
      "source_ids": [
        "bing-public-web",
        "claude-web-search"
      ]
    },
    {
      "heading": "How to distinguish",
      "paragraphs": [
        "Conventional search can return documents without generation. Grounding describes an answer's support in evidence, while RAG describes a way to deliver external context to a model. Product web search may be part of RAG, but links alone are not enough to establish that."
      ],
      "source_ids": [
        "bing-public-web",
        "claude-web-search"
      ]
    },
    {
      "heading": "Example",
      "paragraphs": [
        "In an enterprise assistant, an employee query may retrieve five passages from an internal knowledge base and provide them to the model. If the answer exposes two links, that is a separate interface decision: retrieval may use five passages while attribution shows only two."
      ]
    },
    {
      "heading": "Testing",
      "paragraphs": [
        "Verify the architecture through product documentation, application traces, or a controlled experiment with a known corpus. Record the query, index version, retrieved passages, generation prompt, and answer; without those artifacts, label RAG as a hypothesis."
      ]
    },
    {
      "heading": "Interpretation limit",
      "paragraphs": [
        "RAG reduces dependence on the model's parametric knowledge but does not eliminate irrelevant retrieval, stale corpora, conflicting documents, or generation errors. It also does not guarantee that users will see links to the context used."
      ]
    }
  ],
  "published_at": "2026-09-11",
  "modified_at": "2026-09-11",
  "data_through": "2026-09-11",
  "next_review_at": "2026-10-11",
  "author": "GeoAeoAle Editorial",
  "origin": "editorial",
  "publisher": "GeoAeoAle Editorial",
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "canonical_url": "https://geoaeoale.com/en/dictionary/rag/",
  "claims": [
    {
      "claim_id": "rag-scope",
      "text": "A system should be described as RAG only when retrieved context is confirmed as an input to generation; the answer interface alone does not prove that architecture.",
      "status": "inference",
      "confidence": "medium",
      "source_ids": [
        "bing-public-web",
        "claude-web-search"
      ],
      "publication_status": "public",
      "confidentiality": "public"
    }
  ],
  "sources": [
    {
      "source_id": "bing-public-web",
      "canonical_url": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/guidance/generative-ai-public-websites",
      "title": "Generative answers over public websites",
      "publisher": "Microsoft Learn",
      "source_type": "official",
      "locale": "en",
      "published_at": null,
      "checked_at": "2026-09-11",
      "sha256": "a35496cfe1965d3f1f1862cda5855d6f4825c8316b1ccb5c38722916e94e96d2",
      "license": "Source terms apply",
      "visibility": "public"
    },
    {
      "source_id": "claude-web-search",
      "canonical_url": "https://support.anthropic.com/en/articles/10684626-enabling-and-using-web-search",
      "title": "Enabling and using web search",
      "publisher": "Anthropic Support",
      "source_type": "official",
      "locale": "en",
      "published_at": null,
      "checked_at": "2026-09-11",
      "sha256": "ef0ca674ff3499133c011e53057332bf9acb644c2c4d5b31fe6b836b811c33e9",
      "license": "Source terms apply",
      "visibility": "public"
    }
  ],
  "related_slugs": [
    "ai-visibility",
    "retrieval",
    "source-attribution"
  ],
  "limitations": [
    "A public interface rarely exposes its full retrieval trace, so calling a specific platform RAG should rely on its documentation or observable implementation rather than the appearance of its answers."
  ],
  "corrections": []
}