Promotional language in scientific and institutional documents is usually dismissed as individual failing or passing fashion. Using Bourdieu's field theory, the Sociology of Expectations and Bacchi's problem representation, we argue instead that it reflects scientific documents' structural position and institutional documents' functional mandate in governance. First, an expert-calibrated Large Language Model scores eight promotional dimensions across 268 natural hydrogen documents (1987-mid-2025). Hype, defined as evidential disproportion rather than forward-looking language, occupies the top two of five levels, whereas the models cover the full scale. Two metrics computed on each document's references translate Bourdieu's field theory into network terms, and both behave as it expects. Low out-closeness centrality, the strongest correlate, places those references at the margins of the field. A high hub score shows reliance on canonical authorities, whose symbolic capital the document borrows. Others are high novelty, absent direct measurements, hydrocarbon-system analogies, lower-prestige journals and a techno-economic or environmental focus, together explaining a third of the variance. The register rises as newcomers enter peripheral positions. Second, discourse analysis of 22 institutional documents (2021–2026) shows that register follows mandate. Enactors amplify expectations and Selectors temper them, as in the Sociology of Expectations. Hybrids, the role added here, write expectations into legal categories, funding calls and research agendas with a moderate register. Written into rules, unverified expectations act through procedure, not persuasion. Amplifying texts recur and push caveats to the margins, while tempering texts appear once. This imbalance in information, a constitutive element of procedural energy justice, precedes any participatory process.