Repeated AI relationship-advice chats increased perceived closeness in a five-day study

Researchers recruited 992 participants to seek relationship advice from language models daily for five days. They compared a nonpersonalized model with versions tailored through a survey or conversation memory. They found repeated exposure, rather than personalization, primarily drove several changes in how people interacted with AI, including greater perceived closeness, usefulness and competence. Personalization approaches also produced differences in receptivity to advice and information-sharing attitudes and behavior.