The traditional peer-review system is now under greater pressure. Although the number of scientific manuscripts being sent to journals is still increasing[3, 4], the number of people available to act as reviewers does not keep pace. Researchers are often required to review manuscripts while at the same time dealing with their teaching, their own research, supervision, administrative duties, grant applications, and their publications. Editors also have to cope with the rising number of submissions, find suitable reviewers, assess conflicting reports, pick out any ethical issues, and make fair and defensible decisions with many other duties depending on publishers and journals requirements [5].
This situation raises an important question:
Can artificial intelligence make peer review more efficient, consistent, transparent, and robust without replacing human scientific judgment?
The future of peer review ought not to rest on the notion that machines will take the place of reviewers; rather, it should be built on a thoughtfully arranged partnership in which AI carries out the tasks that involve speed, pattern recognition, comparison, and large-scale screening, and can produce grammatically sound text [2], while human experts continue to be responsible for interpretation, judgment, ethics, and accountability. The aim must not be to develop an “AI reviewer” that recommends acceptance or rejection on its own, since such a method could jeopardize the essential principle of expert judgment and raise serious risks to fairness, accountability, and scientific integrity.
A manuscript is not just a set of words, numbers, tables, and figures; it is the result of a scientific process that must be understood in the context of the discipline in which it exists, its methodology, its ethical considerations, and its social environment. A result based on statistics might seem unusual yet can still be scientifically valid. A methodological decision might appear unorthodox but could suit a given research question. A limitation might be acceptable in one discipline but regarded as unacceptable in another. The decision-making should not be completely handed over to an algorithm. Rather, AI should act as an intelligent component that assists reviewers and editors at every stage of the evaluation process [6-8]. A thoughtfully designed AI-assisted peer-review system could quickly check a manuscript for a wide variety of possible problems and give the reviewer a structured diagnostic report. The human reviewer could then spend more of their time on areas that call for disciplinary expertise, scientific interpretation, and critical judgment.
What AI should not do
One should not allow the enthusiasm about AI to cause us to hand over the fundamental duties of peer review to algorithms.
- AI should not decide on its own whether a manuscript should be accepted or rejected.
- AI must not replace the role of disciplinary expertise.
- AI must not generate a full peer-review report which a reviewer then submits without conducting the required critical evaluation.
- AI must not base a judgment on whether the author has engaged in misconduct or regard statistical unusualness as evidence of data fabrication.
- Uploading of confidential manuscripts to private or public AI systems is never acceptable.
AI use should therefore be supported by secure infrastructure, well-defined data-governance policies, appropriate access controls, and transparent rules on data retention.
Springer Nature could play an important role in developing responsible AI-assisted peer review by creating a secure, transparent, and editor-controlled set of tools integrated into its editorial platforms. Springer Nature listed guidance for researchers, editors, and reviewers” (https://group.springernature.com/gp/group/ai/ai-guidance-for-researchers-editors-reviewers).
Finally, AI should not replace, but assist, human judgment in peer review; responsible collaboration between humans and AI could make scientific evaluation faster, fairer, and more reliable.
References
- Medina, Y.F., et al., A systematic scoping review of essential methodological elements for developing a tool to improve the reporting of consensus studies in classification, diagnostic criteria, and guidelines development. Journal of Multidisciplinary Healthcare, 2024: p. 5813-5830.
- Hosseini, M. and S.P. Horbach, Fighting reviewer fatigue or amplifying bias? Considerations and recommendations for use of ChatGPT and other large language models in scholarly peer review. Research integrity and peer review, 2023. 8(1): p. 4.
- Teixeira da Silva, J.A. and S. Nazarovets, The publish or perish, publish and perish, publish then perish, and now retract and perish cultures in academia. Naunyn-Schmiedeberg's Archives of Pharmacology, 2026. 399(3): p. 3115-3131.
- Mann, S.P., et al., AI and the future of academic peer review. arXiv preprint arXiv:2509.14189, 2025.
- Rees, M., Code of Conduct and Best Practice Guidelines for Journal Editors, 2011. Committee on Publication Ethics (COPE), 2012.
- Kousha, K. and M. Thelwall, Artificial intelligence to support publishing and peer review: A summary and review. Learned Publishing, 2024. 37(1): p. 4-12.
- Checco, A., et al., AI-assisted peer review. Humanities and Social Sciences Communications, 2021. 8(1): p. 25.
- Perlis, R.H., et al., Artificial intelligence in peer review. JAMA, 2025. 334(17): p. 1520-1522.