{
  "dataset": "glossary",
  "record": {
    "id": "grounding",
    "term": "Grounding",
    "category": "knowledge-memory",
    "short_def": "Tying a model's output to verifiable external sources rather than its parametric memory.",
    "long_def": "A grounded answer can cite where each claim came from. Structured data and retrievable content make grounding easier; AI answer engines increasingly cross-check claims against the live page.",
    "see_also": [
      "rag",
      "json-ld",
      "geo"
    ],
    "etymology_origin": {
      "value": null,
      "verify_status": "verify-against-primary-at-build",
      "source_hint": "https://en.wikipedia.org/wiki/Symbol_grounding_problem — 'grounding' derives from the symbol-grounding problem (Harnad, 1990); the LLM 'grounding-to-sources' sense has no single coiner"
    },
    "related_to": [
      "rag",
      "json-ld",
      "geo"
    ],
    "contrast_with": "Unlike RAG, which is the retrieval mechanism, grounding is the property of the output — an answer is grounded when each claim is tied to a verifiable source, however it was retrieved.",
    "example": "AI answer engines such as Perplexity ground responses by citing the live pages they pulled from, letting a reader trace each claim to its source.",
    "source": "https://en.wikipedia.org/wiki/Symbol_grounding_problem",
    "status": "active",
    "why_it_matters": "Grounding is why source-rich, structured, accurate content gets cited; a site that is easy to ground is a site that AI engines quote.",
    "sameAs": [
      "https://en.wikipedia.org/wiki/Symbol_grounding_problem"
    ],
    "bridge_entity": "geo",
    "last_verified": "2026-06-15",
    "md_twin": "/glossary/grounding.md"
  }
}