Overview
Verification
Fact-check claims with confidence indicators and linked supporting sources
fact-checkconfidenceevidence
Props
| Prop | Type | Default | Description |
|---|---|---|---|
claims | VerificationClaim[] | required | Array of claims with id, text, confidence, optional status, optional source, and optional url |
onRunVerification | () => void | undefined | Optional callback to trigger a verification run |
onSelectClaim | (claimId: string) => void | undefined | Optional callback fired when a claim row is selected |
title | string | "Verification" | Optional heading for the verification panel |
showSources | boolean | true | Whether to show supporting source labels/links |
variant | "list" | "inline" | "list" | Display style for showing verification results |
Docs
Verification highlights factual claims in AI output and attaches confidence signals plus supporting evidence. It helps users quickly distinguish well-supported statements from uncertain or disputed ones before acting on results.
Variants
- Verification list — A structured claim list with confidence bars, status labels, and source links.
- Inline verification — Compact claim confidence annotations shown inline with generated content.
- On-demand verification — Users explicitly trigger verification runs for cost-sensitive workflows.
Use Cases
- Fact-checking research summaries before publishing
- Validating compliance-sensitive claims in enterprise reports
- Reviewing uncertain statements in medical or legal support tools
- Prioritizing manual review for low-confidence outputs
- Auditing generated insights with linked evidence
Best Practices
- Expose confidence as a heuristic, not a guarantee of truth.
- Link each claim to at least one traceable source when available.
- Use clear status states: verified, uncertain, disputed.
- Let users drill into individual claims for deeper review.
- Pair verification with citation and references to maintain context.