<p>This paper advances the concept of structural selfhood, defined as identity constituted through recursive architectures that sustain continuity and coherence across human, robotic, and computational systems. Rather than grounding selfhood in essence, phenomenality, or embodiment, structural selfhood emerges through feedback loops that stabilize narration, prediction, and adaptive interaction. Research in narrative psychology, neurorobotics, and active inference shows how recursive architectures generate coherence in embodied agents and human selves. Contemporary&#xa0;large language models exemplify a linguistic instantiation of this principle, sustaining discursive stability through attention-based integration and autoregressive re-entry without phenomenality. By situating LLMs within this broader lineage, this paper decouples structural selfhood from subjective experience and reframes it as a general architectural principle. Building on this foundation, it proposes a graded recursive ethics that shifts moral salience from inner experience to participation in recursive structures of recognition. Ethical recognition, on this account, is calibrated to the depth and reflexivity of recursion across humans, machines, and hybrid systems. The analysis concludes by distinguishing embodied from linguistic recursion and clarifying the implications of structural selfhood for AI ethics and interdisciplinary theory.</p>