How generative AI transforms relational archives into simulations of living persons

Generative AI makes it possible to turn messages, photographs, voice notes, and other platform traces into systems that can converse in the guise of an identifiable person. This critical conceptual essay takes Chinese online discussion of “distilling the ex” as its point of departure. I use “romantic AI resurrection” to describe a specific form of person-specific simulation in which traces from a past relationship are converted into an interactive model of a former partner. The evidence is deliberately limited: public reporting, technical-community walkthroughs, and discussion of ex-partner.skill are treated as diagnostic materials, not as evidence of prevalence. Read alongside work on platformed affect, digital afterlives, AI companionship, relational data governance, and technocultural love, these materials expose a gap in existing accounts. Preparing an archive—choosing messages, removing material, prompting, testing, and revising—is itself an affective practice. The central issue is thus not simply whether a system deceives or comforts its user. It is whether the past expression of a living person can be repurposed as a semiotic partner. I distinguish a right to remember from a right to simulate, and argue for purpose-specific consent, persistent disclosure of synthetic status, friction at import, contestability, and forms of dignified exit.