Peer Review in the Age of AI – What Should Stay Human?

While AI will undoubtedly be able to assist in technical aspects of a review, humans are uniquely able to provide nuanced critique based on their wide-ranging experience on a topic.

We interviewed Laura Dormer, Editor-in-Chief – Learned Publishing, The Association of Learned & Professional Society Publishers (ALPSP) as part of the Peer Review Week 2026 event. Here are her views and perspectives around the current and future state of peer review – and what should stay human. The interview was conducted by Dr. Radhika Vaishnav, Managing Editor, Editage Insights.

Question 1: The Human Element

AI tools are increasingly being used to assist with manuscript screening, language assessment, and even reviewer recommendations. In your view, what aspects of peer review should always remain a human responsibility, regardless of how advanced AI becomes?

AI undoubtedly has a great deal of potential in the scholarly publishing workflow, and it makes sense, given the pressures on peer review, that we look to see how AI might be leveraged here to ease the burden on editors and reviewers. However, I feel (similar to my view on the use of AI by authors), that the element of responsibility itself must always remain with the humans involved in the process. It feels that there is now a move towards ‘humans at the helm’ for the responsible use of AI, and I think this is vital in the case of peer review – humans should set the direction and hold judgement. Editors must remain responsible for the decisions taken on a paper and reviewers must be accountable for the feedback they provide. While AI will undoubtedly be able to assist in technical aspects of a review, humans are uniquely able to provide nuanced critique based on their wide-ranging experience on a topic.

Question 2: Beyond Technical Evaluation

A strong peer review often involves judging novelty, significance, and the potential impact of a study on the field. Do you think AI can ever meaningfully contribute to these assessments, or are they inherently human judgments? Why/why not?

It’s hard for me, as a non-technical expert, to predict where AI will go in the future – it’s already advanced so quickly from when it first came on the scene! But for now, it feels to me that the strengths of AI are on the technical side of review (for example, identifying issues with referencing, or providing statistical support), and looking retrospectively – comparing a submission with what has gone before it. Where it’s use is less helpful, in my view, is in looking to the future, and the potential implications of a piece of work. This is where the experience of human editors and reviewers is so valuable.

Question 3: Trust, Bias, and Accountability

Some argue that AI could make peer review more consistent and less biased, while others worry it could introduce new forms of bias and reduce accountability. How do you see the balance between human and AI involvement in maintaining trust in the peer review process?

Even without AI, we know that peer review is not a perfect process. It’s certainly possible that AI will allow more consistency in review as, on the surface, it should not be susceptible to personal bias in the same way as humans. However, AI tools also have the potential to perpetuate bias already in the system (for example, if they have been trained on historic data that has inherent biases). Again, human judgment will be needed to make sure this is not the case. I also think an important part of the trust in the scholarly record is to know that peer review is not an absolute guarantee, and there must always be the opportunity to reconsider a piece of work, even after publication – be that corrections or, in the most extreme circumstances, retractions – all done in a transparent way. Regardless of the involvement of AI in peer review, the system must continue to work to correct errors identified post-publication.

Question 4: Lessons from Experience

Think about the most valuable peer review feedback you have ever received (or provided)? Could an AI system realistically replicate that kind of contribution?

In my experience as an Editor, what stands out to me is how variable the quality of peer review reports is – some reviews are extremely detailed, while other are more brief (and in some cases, are so cursory that it’s necessary to find a replacement reviewer in order to provide the author with a fair review). So my hope would be that, while I don’t think an AI system could replicate the contribution of the very best, most experienced reviewers, it may be able to assist in improving the consistency of the reviews received overall.

I’d also hope AI is being used to broaden the accessibility of the process.

Question 5: Looking Ahead

Imagine the peer review process ten years from now. What role would you ideally like AI to play, and what role should remain firmly in the hands of human reviewers and editors?

I am reasonably AI-positive, and ideally 10 years from now I would like to see AI playing a supportive role in peer review – streamlining manuscript triage, to avoid unsuitable papers entering an over-burdened peer review system, assisting in the identification and allocation of suitable reviewers, and assisting reviewers and editors to complete their assessment of a manuscript. I’d also hope AI is being used to broaden the accessibility of the process, for example for those whose first language isn’t English. It is important that this takes place within a trusted environment (not simply uploading papers to ChatGPT!), ideally one provided by publishers and embedded in their existing systems to make the process as simple for reviewers as possible, and maintaining the confidentiality of the authors who trust us with their work.

Question 6: “Quick Take” 

If you had to choose just one quality that peer reviewers must preserve in an AI-assisted future, what would it be and why?

The purpose of peer review is to ensure that manuscripts reach readers as the best possible version they can be. Where AI can assist in this – in streamlining processes, removing barriers and providing support – it should be used to do so. But the responsibility for the process should, in my opinion, always ultimately lie with the editors and reviewers involved.