Adrián Fuente-Ballesteros, V. Zuin Zeidler
The fast adoption of large language models has introduced new possibilities and challenges in scientific writing. While artificial intelligence (AI) tools have long supported researchers through grammar checking, reference management, and data processing, recent generative models such as ChatGPT are pretending now to be capable of producing complete sections of academic text that are comparable to conventional manuscripts. This development raises important questions regarding authorship, responsibility, and the integrity of the scientific record. In this work, we propose to examine how AI-generated and AI-assisted text is currently used in analytical chemistry writing, with particular emphasis on recurring linguistic, structural, and bibliographic patterns that may indicate automated drafting in a number of journals' submissions. We propose a reflection on the risks, including superficial, fast, unsustainable, or even misconceptions on science, biased reasoning driven by prompt formulation, and integrity failures linked to AI misuse. Finally, when is the case, we suggest a set of practical indicators and good practices aiming at supporting responsible, transparent, and critically supervised use of AI in analytical chemistry publications.