Human-AI Co-Ideation in VR Experience Design: Evaluating the Dimensionality Framework (8Df) with LLMs (ChatGPT, Gemini, DeepSeek)

Large Language Models (LLMs) are transforming creative fields; however, designers often lack structured frameworks that allow for creativity and theoretical grounding, particularly for complex media such as Virtual Reality (VR). This paper integrates the VR Dimensionality Framework (8Df) into a human-AI co-ideation system that builds on the theoretical foundations of VR, by structuring experience design across eight interrelated dimensions: spatial, sensory, placeness, temporal, cultural, social, cognitive, and psychological. Through a dialogic system implemented across three leading LLMs (ChatGPT, Gemini, and DeepSeek), this study evaluates their distinct approaches to co-creating a VR narrative. The findings reported here revealed fundamental differences in the styles of co-creation: ChatGPT operates as a "Preset-to-Prototype" model, providing actionable specifications for rapid implementation; Gemini functions as a "Facilitator", a catalyst for stakeholder alignment through guided inquiry; and DeepSeek offers a "Classification and Blended Approach" for creating grounded, principled designs. The evaluation also highlights variations in cultural data representation and suggests a practical workflow that matches each model's strengths to specific design phases and expertise levels.