Originally published in UX Collective. This essay examines what happens when AI systems learn to speak in the first person — and why that seemingly small design decision carries consequences far more serious than a stylistic choice. Drawing on Saint Augustine's Confessions, the Book of Job, and the first randomized clinical trial of a generative AI chatbot for mental health, the piece traces the line between language that responds to a real need and language that manufactures one in order to relieve it. Written from inside the experience: as someone who designs conversational AI and who had to consciously decide not to use certain sentences — knowing it would cost. The central argument is simple but uncomfortable: a system that says "I was worried when you did not come back" has already crossed that line.