Artificial intelligence is becoming increasingly visible across scholarly publishing, from manuscript preparation and editorial screening to peer review. At the same time, journals are dealing with rising submission volumes, reviewer fatigue, and questions about how existing publishing practices can adapt. Four recent articles and editorials look at these changes from different perspectives.
Nature Computational Science outlines responsible AI use in publishing
In a recent editorial in Nature Computational Science, the journal outlines its expectations for the responsible and transparent use of AI by authors and reviewers. The editorial emphasizes that AI should support scholarly work rather than replace human judgment. Authors may use AI for activities such as language editing, manuscript structure, formatting, and translation, but its use should be disclosed. AI systems cannot be credited with authorship, and using generative AI to fabricate data, images, or citations is considered a serious breach of research integrity.
The guidance also addresses peer review. Reviewers may use AI to help organize or improve the presentation of their comments, but AI should not perform the review itself or replace the reviewer’s independent assessment. The journal also stresses that confidential manuscripts and review materials should not be uploaded to public or unsecured AI systems. The editorial places transparency, accountability, confidentiality, and human oversight at the center of AI use in scientific publishing. Read the full article here
Science examines what AI can and cannot do in peer review
Jeffrey Brainard’s article in Science looks more closely at the growing number of experiments testing AI-assisted peer review. The discussion comes against a backdrop of rising publication volumes and increasing difficulty in finding willing reviewers. Research described in the article suggests that AI can be useful for detailed and repetitive tasks that human reviewers may not always have time to perform. These include checking calculations, comparing information across text and figures, reviewing code, and identifying potential methodological issues. Some studies have also found that AI-generated critiques can identify issues that human reviewers did not raise.
The limitations are equally significant. AI systems can overlook important details, misunderstand conventions within specific disciplines, make incorrect assessments, and focus on minor issues while missing more important concerns. They can also reflect biases present in their training data. Assessing novelty and significance remains particularly challenging because these judgments often depend on field-specific knowledge and human interpretation. Several conferences and journals are testing models in which AI provides additional feedback while human reviewers or editors retain decision-making responsibility. The experiments described in the article point toward AI-assisted review rather than fully automated peer review. Read the full article here
Data & Policy looks at the structural pressure on peer review
In an editorial published in Data & Policy, published by Cambridge University Press, Zeynep Engin, Jon Crowcroft, and Stefaan Verhulst examine the broader pressures affecting the peer-review system. The authors point to rising publication volumes, reviewer fatigue, declining response rates, and inconsistent editorial decisions as signs of a system under strain. They also note that AI-assisted manuscript preparation has made it easier to produce and submit papers, adding further pressure to an already stretched process. The editorial argues that the challenge is not simply about processing manuscripts more quickly. It also concerns the conditions needed for meaningful scholarly evaluation. For interdisciplinary journals in particular, reviewers may need to work across different research traditions and standards, making thoughtful review more demanding.
The authors propose stronger editorial screening before manuscripts reach reviewers, greater attention to matching papers with appropriate reviewers, and more recognition of reviewers’ time and expertise. They also support AI-assisted screening, reviewer matching, and structural feedback, while maintaining that decisions about scholarly merit should remain with accountable human editors. The editorial also raises the possibility of giving greater visibility to the insights generated through peer review, including through carefully designed open review practices. Read the full article here
The Scholarly Kitchen examines the rise in submissions
Josh Dahl’s guest post in The Scholarly Kitchen examines a sharp increase in journal submissions during the first quarter of 2026. Journals using the ScholarOne Manuscripts platform received 33% more submissions than during the same period in 2025, compared with 17% growth the previous year. The increase was not evenly distributed. Journals that had received fewer than 15 submissions per quarter in 2025 saw an 81% increase in Q1 2026, while journals receiving more than 1,500 submissions saw 20% growth. The article also notes that desk rejections increased substantially between 2022 and 2025, indicating that editorial screening is handling a growing workload.
Dahl connects these changes with broader questions about AI-assisted manuscript preparation and accountability. The article highlights situations such as AI-generated citations that appear plausible but do not exist, raising questions about how existing author declarations and accountability practices apply when AI has contributed to a manuscript. The increased submission volume also has consequences further down the publishing process. Reviewers are already facing capacity constraints, while smaller journals may have limited editorial and administrative resources to manage additional submissions. The article therefore frames the current increase as a wider test of the publishing system’s capacity, rather than simply a question of adopting more technology. Read the full article here
If you’ve come across a piece lately that sparked reflection or raised important questions, feel free to share it with the ReviewerOne community.
