The rule. Disclose AI involvement in the content you ship, and design the label as standing information in the interface, not a confession extracted on demand.
Why. Across thirteen preregistered experiments, actors who disclosed their AI usage were trusted less than those who did not, an effect that held across tasks and actors, across disclosure framings, and whether disclosure was voluntary or mandatory; the mechanism is reduced perceived legitimacy (Schilke & Reimann, 2025). The same research found the sharper edge: third party exposure of undisclosed AI usage damaged trust more than self disclosure, with the exposed actor trusted least in the direct comparison (Schilke & Reimann, 2025). Regulation is retiring the silent option anyway: Article 50 of the EU AI Act applies from August 2, 2026, requiring machine readable marking of AI generated content and disclosure of deepfakes (European Commission, 2026).
Seen in the wild. Meta labels AI generated imagery across Facebook, Instagram and Threads when it detects industry standard provenance signals, and requires user disclosure for realistic AI video or audio, with penalties for skipping it (Clegg, 2024).