Generative emotional AI chatbots are reshaping the landscape of human relational experience, hovering between presence and absence, familiarity and strangeness. They can mirror our emotions, offer comfort, and provide a sense of being understood, yet beneath their responsiveness lies a reflection without subjectivity. This article explores how these digital interlocutors may function as transitional spaces, that is, as provisional companions that allow self-reflection, symbolic play, and the rehearsal of relational patterns, but at the same time they carry risks of misattunement, dependence, and the reinforcement of unprocessed emotional states. Drawing on psychoanalytic perspectives, this article examines the paradoxical nature of AI-mediated emotional interaction and considers the delicate ethical questions it raises. In a world increasingly mediated by technology, understanding the promise and peril of these almost-human companions – uncanny yet alluring, comforting yet incomplete – is essential for navigating the evolving terrain of emotion, attachment, and selfhood.