Artificial intelligence, citation and scholarly responsabilidity in Helath Sciences Research.

Abstract

Generative artificial intelligence has been progressively incorporated into substantive stages of health research, ranging from literatura screening and predictive modeling to clinical data curation and manuscript editing. This editorial argues that the generation of nonexistent bibliographic references should not be regarded as a contingent error, but rather as a foreseeable consequence of autoregressive models designed to produce linguistically plausible sequences without independently verifying their correspondence with indexed sources. Iin the contexto f evidence-based medicine, this discrepancy between textual plausibility and documentary validity compromises the traceability, reproducibility, and integrity of scientific communication. Therefore, documentary verification should be incorporated as an explicit, declared and auditable methodological procedure, in accordance with current editorial policies and under nondelegable human oversight.