Who am I really talking to? Authenticity, transparency, and the case for “AI ticks” in AI-mediated communication

Abstract The growing integration of artificial intelligence (AI) into everyday messaging systems is transforming the conditions under which interpersonal communication takes place. As AI-mediated communication (AI-MC) becomes increasingly seamless and context-sensitive, it blurs the boundary between human expression and algorithmic assistance. This development raises fundamental questions about potential challenges to authenticity in personal relationships and possible strategies to address them. In this regard, we make three main contributions. First, we offer a technically grounded analysis of emerging AI-MC architectures, distinguishing between external prompting, keyboard-level assistance, and fully integrated in-app co-authoring systems. Second, we develop a normative account of authenticity in AI-MC, understood as a second-person, relationship-sensitive notion, drawing on analytic theories of endorsement, relational normativity, and socially situated self-expression. On this basis, we argue that authenticity cannot be directly engineered by design, since it depends on reflective agency and context-dependent interpersonal obligations. Third, we advance a design-oriented proposal aimed at mitigating authenticity-related concerns without resorting to paternalistic restrictions: the introduction of optional and granular disclosure mechanisms (“AI ticks”) that allow users to signal the presence and degree of AI assistance in a context-sensitive manner. We argue that such forms of transparency function not as a substitute for authenticity, but as a pro-ethical condition that enables agents to negotiate authenticity-based obligations within their relationships. The analysis contributes to current debates on the ethics of AI-mediated communication and the conditions of responsible human–AI interaction.