Our research investigates how language conditions direct attention to science in public policy communication, asking whether the language of policy documents affects the visibility of scientific evidence. We approach this question as both an empirical study and a reflection on how public digital documentation systems can foster more inclusive pathways between science and society.

Using the Overton database—a large-scale aggregator of multilingual policy literature—we analyzed 2.25 million policy documents (2009–2024) from Spain, Germany, France, and Sweden. Combining computational text analysis with manual auditing, we measured the prevalence and volume of scientific citations as linguistic traces of direct attention to science. Zero-inflated negative binomial (ZINB) models captured citation sparsity, while multilingual audits validated automated citation detection across languages.

The findings expose that, across countries, English-language policy documents cited science nearly five times as often and forty times as intensively as their counterparts in national languages. However, this pattern varied: while English documents from Germany, France and Sweden were strongly citation-rich, English outputs from Spain showed the opposite tendency, revealing that institutional culture—not language alone—drives attention to science.

Yet across all contexts, a linguistic gap persists: science circulates globally in English, while policy operates multilingually, thereby affecting how public institutions document evidence. Such "linguistic attention gap” matters not only for reliability, but for epistemic justice and social equity. When public records of scientific reasoning in politics are available only in English, non‑English‑speaking citizens, journalists, and practitioners are excluded from full participation in the evidence base that informs policies affecting them.

To close this gap, multilingual public infrastructures—policy repositories and evidence portals—are more than technical supports: they are institutional choices that shape democratic engagement. Building on debates in the digital humanities and the social sciences, we argue that inclusive citation and dissemination systems can narrow the “English advantage” by embedding standardized referencing templates and multilingual editorial workflows. Such infrastructures would not only disseminate scientific evidence across languages but also preserve bibliographic fidelity.

The linguistic attention gap requires intervention at multiple points in the knowledge production chain: from how scientific outputs are published to how policy documents are authored and disseminated. Concretely, this means training the actors who bridge research and policy—librarians, researchers, and policymakers—in publishing accessible, FAIR-compliant data that can circulate across language boundaries. It means developing multilingual citation standards that preserve bibliographic fidelity during translation and adaptation. Extending beyond top-down policymaking, we envision digital infrastructures that see citizens not as information recipients but co-curators of knowledge visibility.

In sum, our work offers both methodological and civic implications: it operationalizes attention to science as a measurable feature of public communication. It reframes the language gap as a shared societal infrastructure challenge. By aligning open data platforms, computational analysis, and community-engaged policy design, we contribute to a vision in which digital scholarship becomes a multilingual public good—ensuring that when policy “speaks science,” it does so in every language of the communities it serves.