This paper examines automated citation workflows as scholarly knowledge infrastructure and asks how automation changes where bibliographic errors and distortions enter, how visible they are, how far they propagate and how readily they can be traced and corrected.
The paper develops a purposive conceptual synthesis of established and recent research, platform documentation, and provider reports concerning citation metadata, reference import, indexing, DOI registration, bibliometric reuse, correction and manipulated or synthetic scholarly content. Cases are used analytically to compare distinct failure modes across workflow stages and infrastructure layers rather than to estimate their prevalence.
While authors remain responsible for checking cited sources, such verification cannot guarantee that a final reference list remains error-free, as defects may be introduced or propagated across connected systems; nor can it ensure that subsequent corrections, retractions or status changes are consistently reflected. The analysis identifies four infrastructural vulnerabilities: variable metadata extraction, propagation across connected tools and databases, platform-specific representation and uneven correction and exposure to manipulation or synthetic content. Operating horizontally across workflow stages and vertically across infrastructure layers, these vulnerabilities can make defective records appear authoritative, obscure their provenance, and hinder attribution and correction.
The paper reframes citation reliability as an infrastructural problem of scholarly records rather than solely a matter of author reference checking. It organises the provenance, propagation, correction, and governance of bibliographic errors and distortions through a two-dimensional analytical framework spanning workflow stages and infrastructure layers.
