Writing by proxy: Authentic authorship in the era of AI-powered communication of scientific knowledge

Abstract

Artificial intelligence (AI) is reshaping biomedical research as both a key analytical tool and an efficient copyeditor. Large language models and related algorithms can help investigators sift through the ever-expanding bodies of scientific literature, connect fragmented biological and clinical datasets, and uncover/recognize meaningful patterns buried within complex multidimensional data. Beyond the bench, these tools can improve and refine manuscript architecture and streamline the flow of information by honing the articulation of simple or complex scientific ideas and sharpening the clarity and cadence of scientific information, which ultimately strengthens the writing in publications and grant proposals alike. For researchers working across languages or at early career stages, these tools allow them to express ideas with greater clarity and fluency, ultimately ensuring that intellectual contributions are assessed on the basis of their conceptual merit and experimental validity. However, the growing persuasiveness of AI-generated writing raises an important concern as to whether persuasive language can create an impression of rigor that is not necessarily matched by rigorous science. AI tools can improve the clarity and presentation of scientific writing, but they cannot establish biological relevance, experimental validation and/or the nuanced judgment required to identify confounding variables, and draw meaningful conclusions. Similarly, an eloquently written grant proposal cannot substitute for an innovative idea, testable hypothesis, feasible experimental approach, and meaningful anticipated contribution. It is therefore essential to view AI tools as instruments of improvement not replacement. They can enhance clarity, efficiency, and readability but they should not replace investigators' intellectual ownership, accountability, and critical interpretation. Disclosure of AI use, together with continued investigators' responsibility for originality, accuracy, citations, data integrity, and interpretation, is essential to maintain trust in scientific discoveries. Similarly, editors, peer reviewers, authors, and funding panels bear a renewed obligation to distinguish writing quality from scientific quality and continue to evaluate approach, innovation, experimental rigor, reproducibility, investigator skills, and scientific and/or clinical significance. Used judiciously and with discernment, AI can free researchers to invest more deeply in the irreplaceable human dimensions of scientific discovery.