This paper examines the interconnectedness of stress-tested banks using financial news coverage. We investigate the behavior of bank networks during periods of stress. We then propose a new measure of systemic risk using text-based eigenvector centrality, a relative metric of influence within a network. We show that this measure provides a valuable complement to existing systemic risk measures. By incorporating soft information into our analysis, we gain a more holistic view of banks’ risk profiles, thus enhancing traditional risk assessment models. Our approach offers a novel tool to study the financial system’s architecture and complements insights from traditional structured data.