As AI tools continue to evolve, they offer new opportunities for scholarly publishing beyond their use in manuscript preparation. This discussion will explore to what extent AI-generated reviews could provide useful support to authors, referees, editors, and readers. We will discuss how AI could complement the human peer-review process and identify which aspects of a manuscript are most suitable for AI-based evaluation, such as completeness, clarity, consistency, reporting standards, or technical checks, while recognizing that scientific assessment and judgment must remain with expert reviewers. Participants are encouraged to share their experiences with existing AI tools, including dedicated manuscript review platforms and experiments with prompting general large language models for manuscript assessment. The goal of the discussion is to develop concrete recommendations for whether, and if so, how AI-generated feedback could be integrated into EGU's transparent publishing model to maintain and strengthen our commitment to innovation, transparency, and the quality and integrity of scientific publishing.