Leading Academic Publishers Update Guidelines on Generative AI for Peer Review
Major scholarly publishers have recently formalized or strengthened their policies governing the use of generative artificial intelligence tools during the peer review process. These updates address growing adoption of tools like large language models while prioritizing confidentiality, human accountability, and research integrity. Elsevier, Springer Nature, Wiley, Taylor & Francis, and others now provide explicit guidance that prohibits uploading unpublished manuscripts to public AI platforms and limits AI assistance to supportive tasks such as language polishing, always with full disclosure and human oversight.
The changes reflect a broader effort to balance efficiency gains from AI against risks including confidentiality breaches, inaccurate or biased outputs, and erosion of the human judgment central to scholarly evaluation. Policies emphasize that reviewers remain fully responsible for their reports, and AI cannot substitute for critical analysis of scientific claims, methods, or novelty.
Context of Rising AI Adoption in Scholarly Workflows
Generative AI tools have seen rapid uptake among researchers for tasks ranging from literature synthesis to drafting text. A December 2025 whitepaper from Frontiers, based on a survey of 1,645 active researchers, found that 53 percent of peer reviewers now use AI tools in their work. Adoption rates are particularly high among early-career researchers at 87 percent, and in regions such as China at 77 percent and Africa at 66 percent. Many reviewers report using AI for drafting reports or summarizing findings, highlighting both opportunities for workload reduction and the need for clearer governance.
This widespread use has prompted publishers to move from ad hoc advice to detailed, enforceable policies. The focus remains on maintaining the confidentiality of submitted manuscripts, which contain unpublished ideas, data, and intellectual property. Uploading any portion to third-party AI services risks violating author rights and data privacy regulations.
Key Elements of Elsevier's Generative AI Policy for Reviewers
Elsevier's updated guidelines state that reviewers should not upload a submitted manuscript or any part of it into an AI tool, as this may violate authors' confidentiality and proprietary rights. Reviewers are responsible for the scientific assessment and judgment of manuscripts. AI tools may only be used in a supportive capacity, such as improving language and structure or assisting with background literature searches, provided confidentiality is maintained and human oversight is exercised. Reviewers must disclose any AI use in their reports, including the tool name and purpose. Basic grammar and spelling checks do not require disclosure. The policy underscores that critical thinking and original assessment in peer review lie outside the current capabilities of generative AI.
Authors using AI in manuscript preparation must provide a dedicated disclosure statement and retain full accountability for accuracy and originality. AI cannot be listed as an author. Similar principles apply to images and figures, with strict rules against fabricating or altering primary research data.
Springer Nature and Wiley Emphasize Transparency and Prohibitions
Springer Nature instructs peer reviewers not to upload manuscripts into generative AI tools due to limitations in accuracy, potential for bias, and risks of nonsensical outputs. If any part of the evaluation draws on AI support, reviewers must declare this transparently. The publisher continues to explore safe, internal AI tools while maintaining strict boundaries on external platforms.
Wiley's October 2025 guidelines reinforce prohibitions on uploading unpublished manuscripts and provide detailed direction on appropriate AI applications for reviewers and editors. The company stresses confidentiality protections and human accountability throughout the review process. Reviewers may use AI for language refinement but must ensure no breach of manuscript confidentiality occurs.
Taylor & Francis Sets Clear Boundaries for Editors and Reviewers
Taylor & Francis explicitly requires that editors and peer reviewers must not upload files, images, or information from unpublished manuscripts into generative AI tools. Failure to comply risks infringing on intellectual property rights. Reviewers must not use AI tools to generate or summarize review reports, though limited assistance with language improvement is permitted under strict conditions. Peer reviewers remain responsible for the accuracy and integrity of their evaluations at all times. The policy aligns with broader commitments to research excellence, transparency, and accountability.
Photo by Nicholas Ng on Unsplash
Evidence of Challenges and Emerging Risks
Despite prohibitions, reports indicate instances of AI-generated peer review content appearing in journals. Some reviews exhibit characteristics of large language model output, such as generic phrasing or fabricated details, raising concerns about quality and reliability. Funding agencies including the National Institutes of Health and National Science Foundation have also prohibited the use of generative AI tools for analyzing or formulating peer review critiques. These developments underscore the enforcement challenges and the importance of clear policies combined with education and oversight mechanisms.
Committee on Publication Ethics guidance reinforces the need for transparency and appropriate disclosure when AI tools are employed. Publishers continue to monitor technological advances and update standards accordingly.
Stakeholder Perspectives Across the Research Ecosystem
Publishers view formalized policies as essential for preserving trust in the scholarly record. Editors and reviewers appreciate guidance that clarifies boundaries while acknowledging potential efficiency benefits from supportive AI use. Authors benefit from consistent expectations around disclosure and accountability. Early-career researchers, who show higher AI adoption rates, express interest in training and equitable access to reliable tools. Institutions and funders are increasingly called upon to integrate AI literacy into research training programs.
Common themes include the irreplaceable role of human expertise in evaluating novelty, methodological rigor, and broader implications of research. AI is positioned as an assistive technology rather than a replacement for expert judgment.
Implications for Research Integrity and Publishing Practices
These policies aim to safeguard the peer review process against risks such as data leakage, biased or hallucinatory outputs, and diminished accountability. By requiring disclosure and human oversight, publishers seek to maintain the credibility of published research. The emphasis on private or approved tools for any permitted AI assistance reflects ongoing efforts to balance innovation with ethical standards.
For the wider academic community, clear policies reduce ambiguity and support consistent application across journals. They also highlight the need for ongoing dialogue among publishers, researchers, and technology developers to refine approaches as capabilities evolve.
Recommendations for Responsible AI Integration
Frontiers' whitepaper outlines practical steps including mandatory transparency around AI use, embedding AI literacy training, strengthening integrity standards, improving data provenance, and ensuring equitable access to trustworthy tools. Publishers encourage reviewers to document any supportive AI use fully and to verify all outputs against primary sources.
Institutions can support compliance through workshops on ethical AI use in research and review processes. Individual researchers are advised to review journal-specific instructions before engaging with AI tools and to prioritize tools that meet high standards of confidentiality and security.
Future Outlook for AI in Scholarly Peer Review
As generative AI capabilities advance, publishers anticipate continued evolution of policies. Exploration of secure, publisher-hosted AI assistants may expand options for reviewers while preserving confidentiality. Emphasis on training, governance, and human-in-the-loop decision making is expected to grow. The goal remains strengthening research quality, reproducibility, and trust in the scientific record through responsible innovation.
These developments signal a maturing approach to AI in academic publishing, one that acknowledges both transformative potential and the enduring value of expert human evaluation.
Resources for Academics and Researchers
Researchers preparing manuscripts or serving as reviewers should consult the latest guidelines from their target journals. Key external resources include detailed policy pages from major publishers that outline disclosure requirements and permitted uses. Staying informed supports compliance and contributes to the integrity of the peer review system.