The scientific peer review system, a cornerstone of academic publishing, is experiencing considerable pressure due to an escalating volume of research papers, a challenge now compounded by the proliferation of AI-assisted manuscript generation. This surge in submissions is overwhelming the volunteer-based review pool, leading to concerns about declining review quality and potential delays in disseminating research findings. Experts across the academic community indicate that the current model is unsustainable without significant adaptation.
The number of academic articles indexed in major databases like Scopus and Web of Science increased by approximately 47% between 2016 and 2022, a growth rate that far outpaces the increase in active researchers. In 2025 alone, over 3.4 million scientific papers were published globally. This substantial increase places an immense burden on editors and reviewers, who often describe peer review as an "unsustainable burden." Many academics report challenging workloads that discourage participation in peer review, a situation that has worsened in recent years.
The advent of large language models (LLMs) and other AI tools has made drafting manuscripts easier, contributing to a rise in submissions, some of which are of lower quality or even fraudulent. Researchers at Cornell University, UC Berkeley Haas, and other institutions examined over 2 million papers uploaded to preprint websites between 2018 and 2024. Their findings suggest that while AI can produce complex writing, AI-assisted papers with more complex prose were less likely to be published in peer-reviewed venues, indicating that sophisticated AI writing can mask weak scientific content. This trend funnels more submissions, including potentially flawed or fabricated work, into the peer review pipeline, increasing the workload for human reviewers.
Before the widespread use of generative AI, the peer review system already faced long-standing weaknesses. These included an over-reliance on voluntary contributions from busy scientists, often leading to principal investigators delegating reviews to junior faculty or students without formal acknowledgment. Issues such as reviewer bias, lack of consistency, and conflicts of interest were also present. AI has not created these problems but has instead exposed and exacerbated them.
AI tools offer potential benefits for enhancing the efficiency of peer review. They can automate administrative tasks such as checking for plagiarism, grammar, and adherence to submission guidelines, thereby allowing human reviewers to concentrate on substantive evaluations. AI systems can also help identify potential reviewer biases and suggest a more balanced pool of reviewers. Furthermore, AI can assist in developing consistent review metrics to ensure more uniform evaluations across submissions. James Zou, a computer scientist at Stanford University, has explored how LLMs can aid scientific peer review, noting AI's value in identifying errors or gaps in research, data, and analysis in early drafts.
However, the integration of AI also introduces critical ethical and methodological questions. AI lacks the nuanced understanding of complex scientific content that human expertise provides, making it challenging for AI to evaluate research novelty and significance. There are risks associated with over-reliance on AI, potential biases in AI algorithms, and concerns regarding transparency, accountability, and data privacy. Editors and reviewers bring a depth of expertise, intuition, and critical thinking that AI cannot fully replicate. The ultimate responsibility for publication decisions must remain with experienced professionals.
To address the growing crisis, clear training and guidelines for editors and reviewers are essential to ensure AI tools are used effectively. Some propose that AI should support and inform human decision-making rather than replacing it. The current situation necessitates proactive development of tools and norms to manage the transition, rather than waiting for the system to collapse under pressure. Rethinking peer review in the AI era involves considering how AI can responsibly complement, but not replace, the critical role of expert reviewers in maintaining the quality and trustworthiness of scientific research.