This paper maps the theoretical and practical challenges of preserving AI-generated content (AIGC) and proposes a coordinated research agenda rather than policy prescriptions. AIGC reveals the networked, processual nature of cultural production, challenging assumptions about human authorship, singular intentionality, and linear transmission. We identify forces that dissolve boundaries—indistinguishability, distributed intentionality, and recursive training loops—examine legal patchworks across jurisdictions, and consider evolving social perceptions that shape cultural recognition. Extending document theory toward evaluation that gives priority to function and drawing on social ontology, we argue that preservation and recognition are mutually constitutive. We outline a coordinated agenda across three fronts: theoretical (models of distributed agency, accounts of extended intentionality, and ethical justification that balances cultural continuity with environmental cost); empirical (ethnographies of human–AI collaboration, longitudinal studies of recognition, and assessments of preservation impact); and technical (metadata with detailed provenance, verification when origins are uncertain or missing, and adaptive selection and sampling, including the preservation of models). We conclude that preservation must expand from storing artifacts to documenting processes and relationships.
Ryo Shiozaki (Mon,) studied this question.