AI-mediated Interpretation of Cultural Heritage: A framing for Chinese collections in European museums

Across Europe, cultural heritage policy increasingly promotes interoperable digital infrastructures and reusable collection data, while museums are beginning to experiment with generative AI for public interpretation. This paper examines how AI can reproduce bias in the mediated interpretation of contested cultural heritage. Focusing on Chinese collections in European museums, it argues that representational asymmetries are not introduced only at the point of AI generation but can already be embedded in catalogue entries and gallery labels, redistributed through platform-level aggregation and reuse, and then re-presented through generative and chatbot-style interfaces that appear fluent, complete, and authoritative. A Yuanmingyuan (the Old Summer Palace) artefact provides the historical and political context for this discussion, while its British Museum catalogue record is used as an illustrative example of how artefact descriptions can foreground dynastic, material, and aesthetic information while leaving acquisition violence and contested ownership outside the main interpretive frame. On this basis, the paper proposes a concise socio-technical framing for understanding how bias travels across description, infrastructure, and AI-mediated interpretation. It further suggests multimodal critical discourse analysis as a useful lens for examining how agency, provenance, uncertainty, and contestation are distributed across these layers. The paper concludes by identifying implications for interpretive transparency, provenance visibility, and institutional accountability in digital museum interpretation.