arXiv:2610.28438v1 Announce Type: new Abstract: The distinctiveness of people with chronic illness as storytellers is how they experience their illness. Symptoms, pain, and biological disruptions are eloquent forms of speech of the body that beget social articulation. The stories people tell about illness shape how suffering is legitimized, and acted upon. Large language models have emerged as new mediators within this landscape---a role once predominantly supported by communities. In doing so, these systems participate in what we term narrative translation: transforming lived illness experiences for particular audiences and purposes. Through 21 prototype-driven semi-structured interviews with people with chronic illness, we contribute an empirical account of how participants navigate this mediation alongside existing sources of support. Our findings surface distinct yet complementary roles participants envisioned for AI and their communities, and how trust in AI was shaped by learning model-boundaries and perceiving understanding. We characterize these findings to inform future designs of AI-mediated narrative support.