A new global reviewer survey from IOP Publishing (IOPP) reveals an increasing attitude towards physical science reviewers regarding the use of generated AI in peer reviews. The study follows a similar study conducted last year, and while some researchers are beginning to embrace AI tools, some are concerned about the potential negative impact, especially when AI is used to evaluate their research.
Currently, IOPP does not allow the use of AI in peer reviews as it cannot meet the ethical, legal and academic criteria that require a generative model. However, there is growing awareness of the possibility of AI supporting rather than exchanging peer review processes.
Important findings:
- 41% of respondents now believe that generative AI will have a positive impact on peer reviews (up 12% since 2024), while 37% consider it negative (up 2%). Only 22%, neutral or uncertain, from 36% last year, show an increase in viewing.
- 32% of researchers already use AI tools to support reviews.
- If the reviewer uses the generated AI, 57% will be unhappy write Peer review reports on manuscripts they co-authored would be unhappy for 42% if AI were used to it.EnhancementPeer review report.
- 42% believe that they can accurately detect AI-written peer review reports on co-authored manuscripts.
Women tend to feel less positive about AI potential compared to men, suggesting sexual differences in AI usefulness in peer reviews. On the other hand, more young researchers appear more optimistic about the benefits of AI compared to their higher-level colleagues who express greater skepticism.
Regarding reviewer behavior and expectations, 32% of respondents reported using AI tools to support them in some way during the peer review process. In particular, more than half (53%) of people using AI said they would apply it in multiple ways. The most common uses (21%) were about grammar editing and improving text flow, with 13% using AI tools to summarize or digest articles being reviewed, raising serious concerns about confidentiality and data privacy. A small minority (2%) allowed to upload the entire manuscript to an AI chatbot.
Interestingly, 42% of researchers believe that they can accurately detect peer review reports of AI-written articles on co-authored manuscripts.
“These findings highlight the need for clearer community standards and transparency regarding the use of generator AI in academic publications. As technology continues to evolve, there must also be a need for frameworks that support ethical and reliable peer reviews.”
“One potential solution is to develop AI tools that are directly integrated into peer review systems, providing support to reviewers and editors without compromising security or research integrity. If these tools are implemented effectively rather than exchanging human judgments, such tools can address ethical concerns as well as reduce risks regarding confidential liability. Add Feetham-Walker.
