Human verification statements under Article 50 of the EU AI Act may raise credibility without improving calibration

Article 50(4) of the EU AI Act introduces transparency obligations for AI-generated or AI-modified public-interest text while allowing visible disclosure to be omitted in specified cases of human review or editorial control combined with editorial responsibility. This raises a question of calibrated trust: does a ‘human-verified’ statement help readers distinguish reliable from unreliable content, or does it simply increase trust regardless of accuracy? In this Comment, we propose the ‘human-verification premium’ as a testable psychological hypothesis, distinguish origin, verification, and accountability as separate cues, and outline how signal detection theory could be used to test whether verification statements improve sensitivity or merely shift the response criterion. We therefore propose testing whether human-verification claims that specify what was checked and who was responsible improve calibration relative to generic verification claims. Their effectiveness should be judged by improved discrimination between accurate and inaccurate content, rather than by higher average credibility. By separating the effect of a verification cue on overall credibility from its effect on accuracy discrimination, the framework makes the psychological consequences of human-verification claims empirically testable and may inform reader-facing disclosure and editorial policy.