Data stories that combine data, visualizations, and prose are widely used for communication, decision making, and persuasion, but producing them typically requires coordinated effort across specialized roles such as analysts, scripters, and designers, which is time consuming and difficult to manage. Existing AI-assisted methods generally treat storytelling as a single-agent task and offer only coarse, global controls, limiting an author’s ability to preserve and shape their communication intention over the course of a narrative. In this work, we present Vistoryteller, a multi-agent authoring system that models the division of labor found in human teams by assigning specialized large language model agents to complementary roles and orchestrating their interactions to generate cohesive, intention-aligned data stories. Vistoryteller supports fine-grained authorial control through two complementary mechanisms: a sketch-based tension-flow control for specifying how thematic emphasis and narrative tension should evolve, and a conversational interface for issuing localized directives to individual agents or to the team. We evaluate Vistoryteller with two controlled experiments and a qualitative user study. Results show that Vistoryteller generates narratives that align more closely with user intentions, preserve coherence across agent contributions, and surface diverse and expressive insights.

Data-driven news articles combine narrative storytelling with data visualizations to inform and influence public opinion on pressing societal issues. These articles often employ persuasive strategies, which are rhetorical techniques in narrative framing, visual rhetoric, or data presentation, to influence audience interpretation and opinion formation regarding information communication. While previous research has examined whether and when data visualizations persuade, the strategic choices made by persuaders remain largely unexplored. Addressing this gap, our work presents a taxonomy of persuasive strategies grounded in psychological theories and expert insights, categorizing 15 strategies across five dimensions: Credibility, Guided Interpretation, Reference-based Framing, Emotional Appeal, and Participation Invitation. To facilitate large-scale analysis, we curated a dataset of 936 data-driven news articles annotated with both persuasive strategies and their perceived effects. Leveraging this corpus, we developed a multimodal, multi-task learning model that jointly predicts the presence of persuasive strategies and their persuasive effects by incorporating both embedded (text and visualization) and explicit (visual narrative and psycholinguistic) features. Our evaluation demonstrates that our model outperforms state-of-the-art baselines in identifying persuasive strategies and measuring their effects.