Spillover Effect of AI-Generated Voice in Video Dissemination and Engagement: Evidence from Bilibili

As demand for online video grows, artificial intelligence (AI) technologies, such as AI generated speech and images, have been used by video creators to improve production efficiency and make profits. However, little is known about the spillover effects of AI tools, particularly AI-generated voice, in video dissemination. We studied the impact of AI-generated voice through an analysis of 129 videos on Bilibili. Applying the panel vector autoregressive model, the results show that AI-generated voice is less effective in promoting video than human voice. Although AI-generated voice improves production efficiency, it decreases video dissemination and audience engagement, which might negatively affect the profit of creators. Further analysis found that this effect reflects two different speech features and people’s differences in preferences for them, providing video creators with a basis to re-evaluate when using AI-generated voice.