Blissful (A)Ignorance: Despite the widespread adoption of AI in communication, people do not suspect AI use in realistic contexts

The growing use of AI in written communication has raised important questions about how people perceive and evaluate others based on their written messages. Past studies have documented a substantial “AI penalty”: communicators are judged more harshly when their use of AI is disclosed or strongly suspected, compared to when their messages are confirmed as human-written. However, under more realistic conditions, audiences may be uncertain, or even completely unaware, of communicators’ potential use of AI. We conducted two online experiments ( N = 1301 ) to systematically examine how both explicit disclosure and uncertainty regarding AI use affect social impressions in realistic communication contexts (e.g., email; social media; texting). In line with previous research, we found that explicit disclosure of AI use resulted in more negative social impressions, compared to confirmed human authorship. However, this AI penalty disappeared or was significantly reduced under more realistic conditions. When potential AI use was not highlighted, recipients exhibited no skepticism and formed impressions that were identical to those of confirmed human writers. Even when potential AI use was highlighted—but remained uncertain—recipients’ impressions were much closer to those of human writers than to those of AI-generated text. Our findings reveal a striking gap between the rapid adoption of AI and social awareness: despite the widespread use of AI in communication, people remain largely unaware of others’ potential AI use, unless this is explicitly disclosed. We discuss how these findings inform ongoing debates about authenticity, trust, and fairness in an era of AI-mediated communication.