The use of AI tools to improve the quality of scientific publications through automated analysis of article preprints DOI: 10.29188/2712-9217-2025-11-3-32-37

57

Introduction. Amid the exponential growth of medical knowledge and publication activity, the speed of disseminating scientific results to the professional community is becoming a critical factor. An assessment of the duration of the peer-review and publication process in biomedical journals indicates that significant delays are often caused not by a lack of scientific novelty, but by formal non-compliance and technical errors in manuscripts.

Objective. The development and substantiation of the implementation of an automated artificial intelligence (AI)-based system for the preliminary analysis of preprints (the Ptolemaea project), aimed at reducing the rate of formal errors and accelerating the publication cycle.

Materials and Methods. The study is based on the principles of Data-Centric AI, scientometrics, and computational linguistics (NLP). The NLP model employed is used not for text generation, but for deep structural and stylistic analysis: detecting violations of IMRAD logic, formatting inconsistencies, verifying statistical indicators, and ensuring compliance with the criteria of specific journals.

Results. An analysis of publication activity shows that the average time from submission to acceptance ranges from 50 to 276 days. Notably, up to 93.2% of errors leading to revisions or rejections originate from the authors (formatting, statistics, incomplete data). The presented technology enables authors to identify these shortcomings at the preprint stage.

Conclusion. Automated preprint analysis is an effective practical tool that lowers the barrier to entry into scientific activity for young scientists and eliminates the routinization of the manuscript preparation process for experienced physician-researchers. The implementation of such systems contributes to improving the quality of scientific communication in healthcare.