Background <p>Within the framework of libertarian paternalism, nudges have been proposed as a way to guide decision-making while preserving individual autonomy, often by exploiting cognitive biases such as framing effects. Although nudges are commonly discussed as a promising means of promoting patients’ best interests, their practical application in real medical contexts remains limited. This study examines whether physicians’ selective phrasing, as a more direct form of intervention than conventional nudging, influences surrogates’ decisions regarding “do not attempt resuscitation” (DNAR) orders.</p> Methods <p>We conducted a series of vignette-based online experiments in which participants evaluated hypothetical end-of-life scenarios. We tested whether subtle linguistic variations in physicians’ explanations—such as mentioning potential harms of cardiopulmonary resuscitation (CPR) or associating DNAR with “naturalness”—affected participants’ consent to DNAR.</p> Results <p>Across studies, participants’ willingness to consent to DNAR was significantly influenced by selective phrasing. Specifically, briefly highlighting the potential harms of CPR and framing DNAR as a form of “natural end-of-life care” were associated with higher rates of DNAR acceptance.</p> Conclusions <p>The findings demonstrate that physicians’ language plays a crucial role in shaping end-of-life decisions without imposing explicit constraints on choice. This suggests that selective phrasing may function as a form of nudging in a broader sense, raising important ethical questions about the boundaries of permissible linguistic selectivity in medical communication.</p>