Rather than treat the surge as a technology problem with a technology fix, we wanted to look at it as a structural challenge, career pressures on researchers, and long-standing underinvestment in editorial infrastructure. I moderated the conversation and was joined by two colleagues who see this from very different vantage points:
- Beth Craanen, Director, Editorial Operations at the American Physical Society (APS), which publishes a concentrated physics portfolio supported by professional editors on staff alongside external academic editors.
- Bob Howard, Executive Vice President, Global Journals at Sage, which manages a large multidisciplinary portfolio that relies heavily on volunteer academic editors.
What the data shows
I opened with findings from our Future of Peer Review 2026 report, which draws on ScholarOne data and industry research. Submissions reached 3.86 million in 2025, up 7% from the prior year, and the first quarter of 2026 came in 33% higher than the same period in 2025.It would be tempting to read that growth as a scientific renaissance, but the desk rejection numbers suggest otherwise. For every accepted manuscript, editors now desk-reject 2.49 others, up from 1.69 in 2022. Meanwhile, the reviewer side of the equation is not keeping pace. The reviewer pool grew to roughly 3 million in 2025, yet fewer than half of invitations are accepted, and it now takes 4.5 invitations to secure a single review, nearly double the 2.3 it took in 2018.
AI adds both acceleration and complication. In surveys, 53% of reviewers report using AI for review tasks, AI text detection tools achieve only about 30% accuracy, and 64% of journals provide no AI guidance for peer review at all, even though most now have policies for authors.
Where the pressure lands, and who absorbs it
The panel confirmed that the aggregate trends are showing up on the ground. Beth and Bob shared that submissions for their journals have risen by more than a quarter this year, with desk rejections climbing alongside them. Just as important, the growth is not uniform. Some titles are absorbing far more of the increase than others, which is why portfolio-level totals can understate the strain on individual journals and their editors. Bob brought the view from Sage's large multidisciplinary portfolio, where volunteer academic editors carry much of the load, offering a useful counterpoint on how scale and editorial structure shape a publisher's ability to respond.We then turned to editorial models. Having worked at small, large, and now medium-sized nonprofit publishers, Beth noted that both the professional and volunteer models have real strengths and trade-offs. At APS, the hybrid approach of staff professional editors working alongside external academic editors gives the team room to focus on innovation, particularly as they evaluate new AI tools for their workflows. Asked what she would build if starting from scratch today, Beth said a hybrid model is the one she would consider.
Writing AI policy that keeps up
With most journals still offering no AI guidance for peer review, I asked the panel why that guidance has been so hard to write. Beth's answer was simple: the technology moves faster than policy can. APS maintains a single policy covering both authors and reviewers, and it revised that policy in June 2026, moving from allowing AI only for light editing to allowing more substantive use with disclosure. Examples include literature synthesis, data analysis, scientific reasoning, figure generation, and translation. APS has also said it will continue to reassess as the tools evolve. Sage also has policies in place, but it's a challenging to continually update them and ensure all stakeholders actually conform to them.Two principles anchor both Sage and APS' approach. Humans remain accountable, meaning authors and reviewers are responsible for the accuracy of anything AI helps them produce. Importantly, human editorial judgment on decisions, along with a clear-eyed choice between quality and speed are absolutely key. It is more about whether you want the work done right or done fast.
Takeaways for the next twelve months
We ended by asking each panelist for the one thing they would want the audience to go back and do in the coming year. Beth's recommendation was to invest in surfacing the right reviewers early in the process and to bring editors along in adopting the tools that make that possible, noting that change management and behavior adoption are often harder than the technology itself. Bob's encouraged the audience to be actively trying new approaches and experimenting. It could be with a subset of the portfolio, but learnings are important when so many changes are happening in our space so fast.That point connects directly to what our data shows about how to relieve the pressure. A few priorities stood out from the session:
- Build reviewer communities where authorship is growing fastest
- Write AI policies for reviewers, not only for authors, and expect to revise them as the technology changes.
- Prioritize relationship-based outreach. Previous reviewers, editorial board connections, and journal community members outperform cold-contact discovery at every volume level.
- Protect editorial judgment by choosing technology that supports editors rather than replacing their decisions.