The Role of Artificial Intelligence in the Lifecycle of Scientific Manuscripts: Authoring, Reviewing, and Editorial Selection

The integration of Large Language Models (LLMs) and Artificial Intelligence (AI) into scientific publishing is accelerating, driven by systemic crises in peer review and academic economic pressures. This manuscript provides a critical three-part analysis: a) AI as a co-authoring tool, balancing its democratizing potential against risks like citation fabrication; b) AI's proficiency in technical review versus its inability to assess novelty, and c) the risk of amplified bias in AI-driven editorial decisions. In the context of peer review, LLMs are being increasingly used for tasks such as preliminary technical verification and language editing. However, their adoption raises critical concerns. Recent evidence confirms that citation hallucination remains a persistent threat. Furthermore, 2025 surveys indicate over 50% of researchers utilize AI during peer review, often violating existing policies. This shift occurs within an exploitative model that relies on unpaid labor while charging substantial Article Processing Charges (APCs). To address these challenges, this paper proposes a framework for sustainable, human-centered integration of AI. A model is proposed in which AI is restricted to technical verification and efficiency, while judgments on scientific merit, ethics, and paradigm-shifting research are reserved for compensated human experts. Nevertheless, the viability of alternative models, including uncompensated hybrid systems, is also discussed_._ Maintaining scientific credibility requires both the ethical integration of AI and a fundamental reform of the economic structures governing research communication.