Crafting Our Own AI Companions: When Fans Blossom Into Virtual Character Designers

Abstract

With the rise of large language models (LLMs), fan creativity is expanding from traditional text-based practices such as Fanfiction to the design of emotionally responsive AI companions. This study investigates how fan designers engage with LLM-based agents as both creative artefacts and relational partners. Drawing on interviews with 20 experienced fan designers, the authors show how fans extend familiar narrative practices, long cultivated in Fanfiction, including character interpretation, worldbuilding and scripting, into the crafting of emotionally resonant AI agents. The authors find that designing these agents is not a one-off technical task but an iterative, affective process shaped by cycles of testing, adjustment, breakdown and repair.

The authors further trace how fan-made agents, initially shared as communal resources within online fandoms, are increasingly appropriated and monetised by commercial platforms, raising concerns about the exploitation of fan labour and the erosion of collective ownership. The authors’ study contributes to fan studies and design research by theorising AI companion design as a fan mode of relational creativity in which affective authorship, community care and platform capitalism are tightly intertwined.