Bureaucratic reputation is now one of the major lenses through which understanding bureaucratic behaviour. However, while the political science literature paved a rich theoretical way for the study of reputation, we do not know what the origins of reputation are, nor we have an externally valid and comparable measure of reputation across agencies and over time. In this paper I address these two gaps in the literature. I propose a structural theory of reputation, for which agencies with a stronger design – that is, endowed with more independence and competences – are more likely to enjoy a good reputation across multiple audiences, regardless of ideological distances between politicians and the agency. Empirically, I leverage a quasi-experimental setting (diff-in-diff) by looking at the impact of structural agency reforms on bureaucratic reputation in the US. By using word embeddings techniques on congressional debates, I also introduce a novel approach to measuring bureaucratic reputation. The preliminary findings suggest that agencies endowed with stronger structures are more likely to build a good reputation, despite the ideological gap between them and Congress. This paper aims to contribute to the literature on bureaucratic politics both methodologically and substantively, bridging structural and reputation-based accounts of bureaucratic behaviour.