Mehmet EmrahKonyaTR
← Back to glossary

Grounding

Temellendirme

D4

Grounding is the practice of connecting an AI system's output or behaviour to supplied or retrievable evidence and context.

Review status: 2026-11-26

Technical explanation

It can make the relationship between an output, a claim, and supporting material inspectable.

Conceptual boundaries

Grounding does not itself prove that an output is true or entailed by a source; sources and citations still require review.

Provider-neutral example

A policy assistant can retrieve approved policy passages, show the passages used for its draft, and direct a reviewer to check whether the draft matches them.

Limitations

A visible source or citation requires separate review for direct support.

Related concepts

Atomic claims and evidence

  1. 1.1NIST AI 600-1, Generative Artificial Intelligence Profile — D4 evidence slice
    Source
    NIST AI 600-1, Generative Artificial Intelligence Profile — D4 evidence slice
    Source role
    Authoritative source
    Exact locator
    MAP 1.1, MP-1.1-001, printed p. 25
    Supported claim
    NIST identifies grounding and retrieval-augmented generation as examples of data-source choices to consider when identifying an intended purpose.
    Last verification
    Review due
    Scope limitation
    This identifies design choices in a risk-management context; it does not define one mandatory grounding implementation.
  2. 2.1NIST AI 600-1, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile — D3 evidence slice
    Source
    NIST AI 600-1, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile — D3 evidence slice
    Source role
    Authoritative source
    Exact locator
    MEASURE 2.5, MS-2.5-003, printed p. 32
    Supported claim
    NIST recommends reviewing and verifying sources and citations in generative-AI outputs as part of measurement.
    Last verification
    Review due
    Scope limitation
    Reviewing sources and citations is a control activity, not proof that every generated claim is entailed or true.
  3. 3.1NIST, Building Evaluation Probes for Agentic AI
    Source
    NIST, Building Evaluation Probes for Agentic AI
    Source role
    Authoritative source
    Exact locator
    Overview; Objectives; Approach — factual-claim/reference-document comparison, structured audit trails, and citation-quality dimensions
    Supported claim
    NIST's evaluation-probe project compares factual claims against reference documents, maps agent decisions to supporting document evidence, and evaluates citations for faithfulness, completeness, and sufficiency.
    Last verification
    Review due
    Scope limitation
    This describes the project's evaluation approach and does not make every displayed citation direct support for a claim.