Large language model-supported companion robots and late-life loneliness: A critical reappraisal of relational AI in cross-cultural care

The growing use of large language model-supported companion robots in later-life care warrants closer scrutiny beyond conversational performance alone. While the original cross-cultural study offers valuable qualitative insights from the United Kingdom and Japan, its findings also expose unresolved questions regarding cross-cultural generalisability, sustained relational value, and the ethical implications of embedding generative AI in companionship technologies. Meaningful support for loneliness in older adults cannot be inferred from dialogue fluency alone, because loneliness is a multidimensional condition shaped by social context, identity, and care environment. This commentary therefore argues for a more context-sensitive and longitudinal research agenda that prioritises adaptive memory, culturally attuned interaction design, and human oversight within hybrid care models. By reframing companion robots as supportive tools rather than substitutes for human relationships, the discussion advances a critical perspective on how relational AI should be developed, evaluated, and governed in psychogeriatric care.