Journalistic Interactive Visualizations (JIVs), such as clickable charts, dynamic infographics, and multimodal stories, constitute not only innovative modes of storytelling but also a distinct form of knowledge, which have barely been studied so far. To fill this lacuna, this paper explores how knowledge is mediated by JIVs, based on a qualitative content analysis of 150 items, sampled from six leading legacy news and financial-news sites (New York Times, Wall Street Journal, Guardian, Economist, Haaretz and Calcalist). The study maps two strategies used by these sites to narrow users’ capacity for polysemous reading of the information: unfitting visualization and naturalizing particular conclusions. These techniques eventually lead to tendentious, sometimes even manipulative, reporting. The paper suggests an initial theoretical framework that enables evaluation of the epistemic standards of quality JIVs, drawing a fine but clear line between essential and acceptable narrowed polysemy, and harmful, tendentious and manipulative forms thereof.