Best Practices for preprint peer review services in the use of AI

This document provides guidance for preprint peer review services on the responsible integration of artificial intelligence (AI) tools, establishing guiding principles and practical steps to support transparency, ethical practice, and research integrity. Life science preprints have increased in adoption since 2013 with a growth of journal-agnostic peer review services since 2017. The use of large language models (LLMs) and other AI tools is rapidly changing research and peer review, raising opportunities for improved efficiency and equity but also risks of fraud, bias, and loss of human judgment. These best practices are designed to encourage responsible use of these tools, without being overly burdensome.

Scope

These best practices are designed for any preprint peer review service, evaluation service or journal organising peer review. Although the guidance focuses on the Life Sciences, it is not exclusive to this field. These particular best practices are not designed for peer reviewers themselves. These best practices and associated guidance may be partially applicable to peer review services that rely exclusively on AI approaches, without human oversight.

Adopting these best practices will help peer review services and potentially traditional journals in ensuring safe, fair, and transparent use of AI in scientific publishing, safeguarding integrity and supporting community trust.