Familiarity, Similarity, Representativeness, and Naturalness: Cues Used to Assess Customized AI News Presenters

Generative AI allows users to control who delivers them news. This study uses think-aloud protocols and in-depth interviews to examine how college students (n = 46) customize AI news presenters and what their selections reveal about preferences in synthetic characters and voices. Participants considered appearance-voice fit and relied on a range of heuristic cues – most notably, whether a presenter looked familiar and/or similar to them, resembled a journalist, looked humanlike, and sounded natural. Social identity theory, the familiarity principle, and the human-likeness-based naturalness concept are used to interpret results. Implications for news organizations and AI companies are discussed.