With the rapid and far-reaching growth of generative artificial intelligence (GenAI) technologies, exemplified by transformative large-scale language systems such as ChatGPT and DeepSeek, the widespread application of large language models (LLMs) has fundamentally reshaped the speed and structural dynamics of academic paper production and journal publishing. At the same time, this technological acceleration has imposed profound and multifaceted challenges upon academic journals, for which the twin imperatives of content authenticity and scholarly originality constitute the very lifeblood. In the context of the continuous iteration of intelligent technologies and the deep integration of algorithms with knowledge production, the boundaries between human writing and machine-assisted generation are becoming increasingly blurred, making the traditional evaluation standards and publication logic of academic journals face unprecedented impacts. This paper, based on a dual analytical lens that integrates frontline publishing practice with a value dialectical perspective, and drawing upon the latest empirical cases and statistical data from both domestic and international publishing sectors, analyzes the current application status of GenAI across the entire workflow of academic journal publishing — encompassing topic selection, manuscript drafting, peer review, copyediting and proofreading, and content dissemination — with a focus on revealing the four core challenges confronting contemporary academic publishing. Specifically, these are: first, the persistent crisis of content authenticity; second, the progressive blurring of academic integrity boundaries; third, the systemic disintegration of trust in peer review; and fourth, the aggravation of reflexive information cocoon effects. Beyond these immediate concerns, the paper further discusses the potential risks of algorithmic bias, data opacity, the instability of model outputs, and the over-reliance on automated systems, which may reshape the knowledge production mechanism and academic evaluation ecology in a subtle but profound way. These systemic risks may also lead to the gradual weakening of independent academic judgment and the progressive homogenization of research expressions across disciplines. The paper further examines how the rapid diffusion of GenAI tools may intensify the structural imbalance of academic resources across regions and institutions, thereby undermining the fairness and inclusiveness of scholarly communication on a global scale. With an aim to transform academic journal publishing from a passive gatekeeper to an active intelligent publishing guide, this paper proposes a set of actionable governance strategies structured around five pillars: first, the improvement of editorial mechanisms with deep intervention in the whole publication process; second, the systematic strengthening of editors' professional judgment and AI-related technical literacy; third, the construction of academic norms and publishing ethics systems specifically oriented to the intelligent era; fourth, the implementation of human–AI combined review and detection protocols designed to ensure both operational efficiency and epistemic reliability; and fifth, the cultivation of a diversified and open academic ecosystem that actively encourages transparency, cross-institutional collaboration, and responsible innovation. Collectively, these strategies aim to promote the academic community in establishing a new form of trust contract premised on productive human–AI coexistence. Furthermore, it emphasizes the necessity of building cross-institutional governance frameworks and dynamic regulatory mechanisms, so as to continuously adapt to technological evolution and to maintain the core values of academic publishing. In this way, it seeks to provide a reference scheme with theoretical depth and practical operability for academic journals to cope with the publishing changes in the intelligent era and to realize the coordinated development of technological empowerment and value adherence.