With the deepening involvement of generative artificial intelligence in literary creation, prompts have become a key factor shaping the effectiveness of human-AI collaborative writing.This study investigates how prompts stimulate AI's literary creativity by conducting generation experiments on the DeepSeek model using five types of prompts, and by employing a combined subjective-objective evaluation framework for both quantitative and qualitative analysis.Through a mechanism of“semantic orientation-cognitive-rhetorical intervention, ”prompts translate human intentions into executable creative operations for AI.Abstract-concept prompts activate cross-domain analogy, while integrative prompts facilitate the synthesis of heterogeneous semantics.The cultural parasitism of AI not only reflects the collective unconscious of humankind, but also creates opportunities for defamiliarized expression, thereby fostering the emergence of distributed creativity.The findings highlight the crucial role of prompt design and multi-round generation strategies in AI-assisted literary creation, and provide theoretical and methodological reference points for human-AI collaborative writing in the context of digital humanities.