Enhancing national soft power has long been a vital strategic task for China. With the continuous innovation and widespread application of AI, the pathways of information dissemination and cognitive formation are being rapidly reshaped, and the paradigms of international communication and national soft power construction are undergoing profound transformation. In light of these shifts, it is necessary to incorporate AI as an emerging medium into a key dimension of national soft power analysis. This study constructs an analytical framework based on AI attention allocation, quantitatively evaluates the explicit articulations of mainstream large language models regarding national soft power, and measures the actual state of national soft power across countries in the era of intelligent communication. The findings show that national soft power in this era exhibits a pronounced Matthew Effect, in which strong countries grow stronger while peripheral ones are further marginalized, and the soft power of leading countries remains relatively stable across multilingual environments. Different AI models also vary considerably in their attention allocation, showing a discernible native-language advantage. Against this backdrop, China should promote a more balanced national soft power structure in the era of intelligent communication. By actively participating in international AI governance and cooperation and strengthening the development of multilingual data ecosystems, China can contribute to a more equitable and inclusive global narrative as technology evolves, and help prevent AI from becoming a game for wealthy countries and the rich.