Algorithmic reputation: toward a theory of the construct in the era of generative models

Working paper — conceptual anchor of the research program on algorithmic reputation. The empirical companion piece (10.5281/zenodo.18802347) develops the exploratory application to AI search engines. This article introduces the construct of algorithmic reputation and argues for its theoretical autonomy. It is defined as the representation of an entity that is constructed and reproduced by an algorithmic mediator, and is thus distinct from both the aggregate human judgment studied by corporate reputation research (Fombrun, 1996; Ponzi et al., 2011) and the message–source dyad examined in source credibility studies (Hovland & Weiss, 1951; Pornpitakpan, 2004). The shift in evaluator—from the human stakeholder to the statistical model—renders neither of these classical constructs isomorphic to the new object of study. The contribution is fourfold. First, it traces the conceptual genealogy of algorithmic authority as a direct antecedent (Gillespie, 2014; Caplan & boyd, 2018; Bucher, 2018; Diakopoulos, 2016) and connects this tradition to large language models. Second, it proposes two analytically distinguishable dimensions—entity authority and entity credibility—and explicitly confronts them with the bidimensional model of organizational reputation by Rindova et al. (2005). Third, it delineates the boundaries of the construct vis-à-vis corporate reputation and source credibility, and formulates three falsifiable propositions that specify the conditions under which it could be refuted. Fourth, it maps three modalities of the construct—synthetic, search-engine, and platform reputation—with the synthetic modality as the object of a subsequent contribution currently in preparation, while the platform modality is incorporated as a boundary case whose full inclusion within the umbrella construct is deliberately left open.