<p>Personalized medical care increasingly incorporates algorithm-driven systems. In contemporary diabetes management, many affected individuals regulate blood glucose levels in collaboration with automated insulin delivery (AID) systems. These devices interact closely with the body, producing data, emitting alerts and requiring constant interpretation and maintenance. Although the devices cannot literally “speak” by articulating words, users often perceive them as interactive agents that participate in the task of managing glucose levels. Language becomes central in how users narrate and negotiate these interactions. Rooted in Bakhtin’s concept of voice and post-humanist theories of language, the study conceptualizes language as a relational, distributed, and co-constitutive process across human and non-human actors. It analytically&#xa0;draws on discourse analysis with particular attention to linguistic qualities of voicing sequences. The analysis focuses on constructed dialogues from 31 qualitative interviews with users of algorithm-driven diabetes technology based in Germany to examine how speakers linguistically animate their devices and negotiate intersubjective relations with them. These sequences show how users perform their devices in constructed dialogues: what they have them “say” and how they linguistically shape their “voices”. First, the analysis reveals how algorithm-driven devices are linguistically constructed as (unequal) partners in meaning-making processes related to complex diabetes management. Second, the findings illuminate a broader human tendency to imagine objects as animated. They show how the voices of otherwise voiceless machines are enacted, embodied, and integrated into complex relational frameworks. Overall, the analysis demonstrates how human and machine voices are discursively positioned within socially meaningful human–machine relations, highlighting the role of language in shaping relational asymmetries and imagined co-agency.</p>