Personalization and social closeness buffer negative reactions to AI-generated birthday wishes on social media: evidence from two experiments

Generative AI is reshaping how friendships are maintained, yet the implications of this technology for such social bonds in online contexts remain unclear. We conducted two 2 × 2 factorial between-subjects experiments to investigate this issue. In Study 1 (N = 341), we manipulated the source of a greeting message (human-written vs. AI-generated) and text personalization (non-personalized vs. personalized). In Study 2 (N = 325), we manipulated message source (human vs. AI) and the closeness of the friend (distant vs. close friend). Across both studies, AI-generated wishes were perceived as less sincere and as requiring less effort than human-generated wishes. As a result, recipients liked the AI-generated wishes less and reported lower friendship satisfaction. Additionally, personalization mitigated AI’s drawbacks, restoring perceived sincerity and effort; closeness buffered AI’s harm, so close friends’ AI messages were judged less negatively. These findings refine gift-giving theory and pinpoint when AI-generated messages undermine or sustain interpersonal bonds.