Subject Matter Experts (SMEs) are essential in serious games for cultural heritage, where their input ensures accuracy and ethical integrity. Designers must collaborate closely with SMEs to encode cultural content into game components, often relying on structured frameworks to support and formalize this process. However, such collaboration is time-consuming, costly, and often out of reach for small or underfunded cultural projects—especially those focused on niche topics with limited audiences. This study explores the extent to which generative AI can reduce this dependency. Using ChatGPT within the Co.Lab framework, we developed Edgar à GoGo, a prototype inspired by Varèse’s Poème Électronique, combining AI-generated content with a human-authored game structure. The experiment examines how general knowledge, enhanced by selected project-specific materials, can partially substitute the traditional SME–designer dialogue. The prototype was reviewed by three SMEs familiar with the heritage item, offering insights into the coherence and completeness of the AI output. Findings suggest a hybrid workflow with AI aiding human design. While not a replacement for expert insight, generative AI may serve as a catalyst for ideation and discussion, especially in low-resource settings.