From Failing Forward to Projecting Forward? Anticipatory Logic in EU Crisis Communication (2008–2025)

The European Union’s development is widely understood as crisis-shaped, from the global financial and sovereign debt crises to migration and COVID-19, and increasingly also by foreign interference and other hybrid threats, such that crisis has increasingly become a persistent condition of EU governance (“crisification”). Within this debate, the influential “failing forward” account depicts integration as fundamentally reactive: initial crisis deals tend to be minimalist and leave key problems unresolved, which later reappear as renewed pressure for further institutional reform (Jones et al. 2016; 2021). Yet, alongside this reactive pattern, EU policymaking has been accompanied by growing investment in preparedness, resilience, and strategic foresight, raising the possibility that crisis-related integration is increasingly pursued prospectively and aimed at preventing potential future breakdowns rather than responding to present failures. This article advances and empirically explores the proposition that EU institutions increasingly engage in “projecting forward”: a forward-looking crisis logic in which anticipated or potential crises are invoked to legitimate measures oriented toward risk prevention, crisis preparedness, resilience, and future-proofing. Theoretically grounded in constructivist and discursive institutionalist perspectives (Schmidt 2008), the article treats institutional discourse as a key site where crisis logics are both reflected and co-constituted. While anticipatory framing does not automatically equate to anticipatory integration, tracing how institutions justify measures, reactively versus anticipatorily, provides a valuable window into the evolution of EU crisis politics and legitimation practices. Empirically, the study analyzes a longitudinal corpus of EU institutional crisis communication from 2008–2025. Methodologically, it develops a scalable framework that combines LLM-based classification with topic modelling and temporal analysis. First, LLMs are used to identify and differentiate reactive versus anticipatory legitimation patterns and to capture latent, context-dependent framings that are difficult to operationalize with keyword-based approaches. Second, topic modelling maps how anticipatory repertoires cluster by crisis type, and how they intensify or decline over time. The article contributes to theories of crisis-driven integration by proposing that projecting forward may operate alongside reactive “failing forward” dynamics. While treating discourse as a proxy cannot fully demonstrate that integration outcomes follow from these framings, it opens the door for further research beyond discourse to assess whether a new crisis-driven integration logic is emerging. Methodologically, it shows how LLM-enabled approaches can make complex, latent crisis framings observable and comparable over time in EU institutional discourse.