More than ever, AI has the power to recommend, suggest, modify, and generate messages on behalf of human communicators. This is a phenomenon known as AI-mediated communication (AI-MC). While AI-MC can enhance communication efficiency, it also influences outcome perceptions, language use, trust, and attitudes in ways not yet fully understood. Over the last six years, Naaman's research has explored these effects, revealing how AI-MC impacts message ownership, interpersonal evaluations, and even shifts in expressed attitudes. AI-MC also raises ethical concerns, disproportionately affecting different demographic groups and their AI usage perceptions. This talk highlights experimental findings, emphasizing the need for responsible development to address the profound implications of AI-MC on human communication.