<p>Against the backdrop of digital technology reshaping cultural experiences, museums and other knowledge institutions increasingly adopt artificial intelligence (AI) for public engagement, with AI-driven virtual digital humans introduced as docents in high-immersion environments such as holographic cabins. To enhance visitors’ sense of immersion, museums have widely introduced these virtual humans, yet how their two core media attributes—anthropomorphism and intelligence—jointly influence visitors’ psychological processes, particularly trust and spatial presence, remains unclear. Through an empirical study of 351 visitors in the intelligent interactive holographic cabin at Jilin Provincial Museum in China, this study reveals the dual-pathway mechanism by which AI-generated virtual digital humans’ mediality affects continuous participation intention. The findings reveal a context-specific boundary condition of Media Equation Theory: in knowledge-intensive museum settings, the degree of anthropomorphism in AI-generated virtual digital humans does not significantly enhance visitors’ sense of trust; rather, their level of intelligence serves as the key driver for establishing trust. Meanwhile, although anthropomorphism fails to build trust, it significantly enhances visitors’ sense of “being there,” namely spatial presence, and indirectly drives continuous participation through this experiential pathway. These findings refine the prevalent understanding of anthropomorphism in human–computer interaction and contribute to the algorithmic-trust literature, indicating that future AI design in cultural heritage should focus on deepening knowledge capabilities rather than merely pursuing visual realism—an insight with broader relevance for all fields aiming to disseminate knowledge and education through AI.</p>