Mapping the Public Discourse on Generative AI: A Longitudinal Topic Analysis of Microblog Texts

The rapid proliferation of Generative AI has triggered intense public discourse on social media platforms. However, the longitudinal evolution and thematic dynamics of this online discourse remain underexplored. This study addresses this gap by analyzing 5,499 microblog posts related to “Generative AI” from Sina Weibo between April 2022 and March 2024. We employed a Latent Dirichlet Allocation (LDA) model to identify latent topics and constructed a topic heat evaluation index, integrating topic strength and user participation, to track the prominence of these topics over time. Our analysis first identified three overarching themes: (1) technical applications and tool development, (2) the developmental landscape in China, and (3) impacts on daily life. More significantly, the longitudinal analysis of topic heat reveals a distinct four-stage evolutionary trajectory of public focus: from initial concerns about misuse (e.g., academic cheating) and the capabilities of core technologies (e.g., GANs), to a concentrated fascination with specific applications like ChatGPT and their business potential, and ultimately culminating in discussions around legal and regulatory frameworks. This research provides a quantitative framework for mapping the dynamic evolution of public opinion on disruptive technologies and offers empirical insights for policymakers and industry stakeholders monitoring the societal integration of AI.