In safe hands: Human-led publishing and where AI can help
Authors: Melanie Dankel1, Mara Lambert1, Tara Crandon1, Hien Ho1
1. JBI, Adelaide University, Adelaide, SA, Australia
Spotting artificial intelligence (AI) in manuscripts has become a bit of a sport in recent times, with conversations between editors and readers verging on a ‘did they or didn’t they?’ rhetoric. The internet is littered with examples of AI fails, and even the humble em dash has been demonised thanks to its association with AI output.
For a journal with both a global authorship and a global readership, it’s a fine line to tread between embracing the benefits of AI in producing high-quality content and avoiding the pitfalls (and retractions!) of AI slop. While guidance from bodies such as COPE and individual journal policies go some way to ensuring rigorous standards, the reality is that every journal must operate on a system of trust with its authors, editors, and peer reviewers. In this blog, we explore some of the benefits and challenges of using AI in a peer-reviewed journal.
The role of AI in peer review
JBI Evidence Synthesis has grown exponentially in recent years in both submissions and subscriptions. It received its first impact factor of 2.7 in 2023, and this currently sits at a healthy 6.8. This increase in reach has been accompanied by a large increase in submissions to the journal, which has placed pressure on the journal office to manage the volume of submissions. Even with a rejection rate of approximately 70%, the journal now requires more peer reviewers to ensure that we continue to publish relevant, high-quality, trustworthy evidence syntheses.
While AI presents numerous opportunities, which we will discuss later in this article, it also has its challenges. One of these challenges is when peer reviewers upload manuscripts into AI and ask it to complete the peer review. Not only does this violate copyright, but it also results in a poor-quality peer review that has very little value. One author stated that the feedback they received was ‘vague, confusing and did nothing to improve my work’, going as far to say that AI ‘undermined’ peer review. When we suspect peer reviewers have used AI to generate their comments (rather than to refine their language), our editorial team steps in to carefully review the manuscript and all peer review comments. We also keep a record of any such instances and de-list the peer reviewer.
Peer reviewers are alerted to the prohibition of uploading manuscripts into AI in the letter inviting them to review a manuscript. They are allowed to upload their own comments to AI for refinement in grammar and expression, which can be helpful for authors whose primary language is not English.
Setting the standard for AI in evidence synthesis
While we can acknowledge the opportunities for AI to assist in evidence synthesis for tasks such as data extraction and deduplication, it is important for authors to remain in the driver’s seat. JBI recently published a joint position statement on AI with Cochrane, the Campbell Collaboration, and the Collaboration for Environmental Evidence. One of the key messages of this statement is ‘human oversight’. Another key message is authors’ responsibility for any AI output, and this includes the decision to use AI.
JBI Evidence Synthesis asks authors to declare any AI use, including the use of AI to refine grammar, punctuation or expression. Where authors intend to use AI at any stage in the conduct or reporting of evidence synthesis, this must be declared in the methods, and specific details of how it will be used appended.
When it comes to articles in languages other than English, JBI recommends that authors do not limit their search strategy by language in order to avoid language bias and to ensure all relevant sources are included in the synthesis. This often means that authors rely on DeepL (an AI-powered language platform) or similar software to translate those sources into their primary language. We ask authors to indicate whether they will verify those translations (ie, via human translators) to ensure the quality and accuracy of the data. This, once again, emphasises a human-first approach to AI.
Another important factor to note is that JBI Evidence Synthesis does not consider AI to meet the requirements for authorship. If content is produced by AI that breaches ethical or legal standards, the authors retain responsibility for those breaches. As mentioned previously, an important consideration is the breach of copyright in uploading manuscripts into AI systems. All authors complete a transfer of copyright on submission of a manuscript, unless they choose to publish open access. Uploading a manuscript essentially creates an unauthorised digital copy of the manuscript. In addition, large language models learn from the uploaded data, which many creators argue breaches their copyright where neither permission nor compensation is addressed.
Opportunities in the publishing landscape
The story is not all doom and gloom. AI does offer opportunities for authors whose primary language is not English, especially in helping improve equity in publication. We have previously published an article on the dominance of English within scholarly publication, and the disadvantages authors with other primary languages face, as do the communities their research pertains to. AI offers an opportunity for these authors to be able to disseminate their research rather than face journal rejection due to poor English expression.
Closer to home, the steady increase in submissions to our journal means our receiving editors are dealing with an ever-expanding workload. We are currently exploring whether AI can assist with time-consuming administrative screening tasks, such as checking that manuscripts meet basic submission requirements. Our intention is to see whether AI might streamline our administrative processes so that peer review proceeds more efficiently, and so that receiving editors have more bandwidth to do what they do best: comprehensively review manuscript content. Being within the Adelaide University ecosystem allows us to use a closed AI platform that removes the risk of breaching copyright or of the data being scraped by AI. It is important to us that humans are kept in the loop at all stages of the editorial process, even administrative ones. If we are to consistently use AI in our workflows, all tasks will be checked by a human.
Evidence syntheses and protocols in the journal are now published with an AI-generated ‘In brief’ plain language summary, which is added to the review at publication by the publisher. Increasing the accessibility of scholarly research makes it easier for that research to be communicated to a broader audience. Authors have the option of withdrawing this summary if they feel it has errors or does not represent their research accurately. To date, only one author has requested this.
The future is uncertain but hopeful
The capabilities of AI and the best-practice guidance around its use are evolving at a rapid pace. While this tool opens the door to a world of possibilities, we would urge users, including authors and peer reviewers, to not forget that ‘Creating is the fun part’. Our hope is that we can continue to operate on a basis of trust with our authors, peer reviewers and editors, which will, in turn, ensure trust in the output of the research we publish.
References
Barker, T. H. (2025, March 10). ‘Vague, confusing, and did nothing to improve my work’: How AI can undermine peer review. The Conversation. https://theconversation.com/vague-confusing-and-did-nothing-to-improve-my-work-how-ai-can-undermine-peer-review-251040
COPE. (n.d.). COPE focus in artificial intelligence. https://publicationethics.org/cope-focus/artificial-intelligence
Csutoras, B. (2025, April 29). The em dash dilemma: How a punctuation mark became AI’s stubborn signature. Medium. https://medium.com/@brentcsutoras/the-em-dash-dilemma-how-a-punctuation-mark-became-ais-stubborn-signature-684fbcc9f559
Dankel, M., & Lambert, M. (2023). Abstracts for a wider audience: Promoting linguistic diversity in JBI Evidence Synthesis. JBI Evidence Synthesis, 21(5), 833–834.
Flemyng, E., Noel-Storr, A., Macura, B., Gartlehner, G., Thomas, J., Meerpohl, J. J., Jordan, Z., Minx, J., Eisele-Metzger, A., Hamel, C., Jemioło, P., Porritt, K., & Grainger, M. (2025). Position statement on artificial intelligence (AI) use in evidence synthesis across Cochrane, the Campbell Collaboration, JBI, and the Collaboration for Environmental Evidence 2025. JBI Evidence Synthesis, 23(11), 2162–2166.
Hines, S., & Francis, E. (2026). Submission declined: An analysis of reasons for rejection in JBI Evidence Synthesis. JBI Evidence Synthesis, 24(5), 852–859.
JBI Evidence Synthesis. (2026). Information for authors. https://edmgr.ovid.com/jbisrir/accounts/ifauth.htm#TOC6
Loeppky, J. (2024, June 16). 32 times artificial intelligence got it catastrophically wrong. Live Science. https://www.livescience.com/technology/artificial-intelligence/32-times-artificial-intelligence-got-it-catastrophically-wrong
Mineo, L. (2026, May 27). Funny but serious, Chieng issues an AI warning to grads. The Harvard Gazette. https://news.harvard.edu/gazette/story/2026/05/funny-but-serious-chieng-issues-an-ai-warning-to-grads/
Nussbaumer-Streit, B., Streit, M., & Porritt, K. (2026). Leveraging artificial intelligence for evidence synthesis. JBI Evidence Synthesis, 24(7), 1280–1282.
Sridharan, K., & Sivaramakrishnan, G. (2026). Artificial intelligence in the retraction spotlight: Trends, causes and consequences of withdrawn AI literature through a systematic bibliometric review. Frontiers in Research Metrics and Analytics, 10, 1737168.
Thorpe, L. (2025, September 9). No consent and no compensation: How AI treats your copyrighted work. Digital Rights Watch. https://digitalrightswatch.org.au/2025/09/09/no-consent-and-no-compensation-how-ai-treats-your-copyrighted-work/
To link to this article - DOI: https://doi.org/10.70253/FKWR2887
Declarations
This blog was written entirely by humans.
Disclaimer
The views expressed in this World EBHC Day Blog, as well as any errors or omissions, are the sole responsibility of the author and do not represent the views of the World EBHC Day Steering Committee, Official Partners or Sponsors; nor does it imply endorsement by the aforementioned parties.
