How Scientists Can Use AI Responsibly in Research and Publishing

Learn how scientists can use AI responsibly for data analysis, figures, manuscript writing and peer review while protecting research integrity.

For scientists, artificial intelligence (AI) has the potential to improve efficiency, streamline tasks and polish communications. The payoff is more time for creative thinking, critical analysis and scientific judgment that only humans can provide. Researchers are embracing these tools. A survey of 6,000 German researchers found that more than 50% are using AI in their daily work. When it comes to publishing, how can researchers use AI effectively and ethically?

Many AI tools can accelerate the processing of large data sets and support statistical modeling. By comparing multiple variables and conditions quickly, these tools can yield faster results and suggest new areas of investigation. If you’re designing a custom AI tool specific to your lab, testing and validation are essential. Researchers must understand how the model was trained and where its limitations lie. Authors are responsible for scrutinizing and verifying all data outputs and disclosing them in the methods section of the manuscript. Ethical standards still apply to AI data processes and outputs, so rigor and reproducibility remain essential standards.

A paper’s figures are another area where caution is key. Visuals often provide the first impression of a study’s findings, making accuracy critical. Graphs should be based on real source data, rather than from generative image software. APS journals allow the use of AI to create schematics and conceptual illustrations, but the authors should verify the accuracy of any resulting images.

Authors should also be mindful of copyright. Copyrighted images should not be used as AI inputs without permission from the copyright holder. Regardless of how a figure is created, authors are responsible for its accuracy and appropriate use. Any use of AI-generated images should be disclosed in the acknowledgments and figure legends.

For many researchers, writing and revising manuscripts is a significant time investment. AI can be useful as a second set of eyes, helping improve organization, clarity and flow of content. APS journals permit the use of AI tools for this use if it is disclosed.

However, drafting a manuscript completely with generative AI presents risks. These tools can produce inaccurate information, fabricated statements and made-up references. Researchers should carefully review AI-generated suggestions before incorporating them into a manuscript. Any use of AI writing assistance should be disclosed in the acknowledgments.

The role of AI in the peer review process is complicated. At the heart of peer review is a commitment to confidentiality. Reviewers and editors are entrusted with unpublished work, and uploading manuscripts to external AI systems that may store, retain or redistribute material breaks this trust.

Peer review also depends on the expertise and experience of scientists. AI cannot replace the critical evaluation that peer review requires. APS journals allow AI for language-support functions only when confidential manuscript content is not disclosed to external AI systems.

AI is becoming a routine part of research and publishing. The uses of AI are constantly evolving, and APS Publications continues to evaluate its policies. For the latest guidance on using AI responsibly in APS journals, visit journals.physiology.org/editorial-policies#ai-policy.

This article was originally published in the November 2026 issue of The Physiologist Magazine. Copyright © 2026 by the American Physiological Society. Send questions or comments to tphysmag@physiology.org.