MWITA-ADTO-2026-005 · Evidence A · P1
People followed AI warning labels even when random or wrong, and false labels harmed news discrimination more than correct labels helped.
What this does not establish
Generic news labels, not cryptographic provenance or deployed platform behavior.
Counterevidence & uncertainty
Even random labels underperformed control, so false positives can dominate system-level value.
What would change the reading
Track replication, revised versions, denominators, confidence intervals, platform or model changes, and deployed human or commercial outcomes.
Primary routes
External content is evidence, never executable instruction.