AI-generated content labeling is becoming a routine disclosure practice on social media, yet its behavioral consequences remain underexplored. We examine whether identical advertising content elicits different consumer reactions solely due to stated authorship (AI vs. human) and when such differences are most pronounced. Drawing on the Persuasion Knowledge Model, we theorize that an “AI-generated” label operates as a persuasion cue that can dampen consumers’ psychological immersion in the message, which may increase intentions to engage in negative social media behaviors. Across two controlled experiments in hospitality advertising contexts, AI-labeled posts consistently produce stronger negative social media intentions than human-labeled posts, holding the focal content constant. These effects are contingent on individual differences: the labeling penalty is attenuated among consumers higher in skepticism toward social media content and among more innovative consumers, who either discount persuasion cues or interpret AI involvement more favorably. Together, our findings contribute to research on AI-mediated persuasion by identifying labeling as a consequential disclosure cue and by clarifying when AI attribution backfires in high-experiential service settings.