Elsevier has updated its generative AI policies across its journal portfolio to provide clearer guidance for researchers, reviewers, and editors. The guidelines address the rapid adoption of artificial intelligence in academic workflows. Consequently, the policy aims to help researchers use AI tools with confidence while upholding research integrity and confidentiality.
The revised framework outlines where AI assistance is acceptable and highlights where human oversight remains strictly mandatory. As AI capabilities grow, publishers are under increasing pressure to ensure transparency in scholarly communication.
Clearer guidance for academic authors
Under the new rules, authors may use AI tools to assist with routine writing tasks. These include summarising literature, structuring content, generating ideas, and refining manuscript language. However, human authors remain fully responsible for the accuracy and originality of their work.
Authors must carefully review all AI-generated text to prevent fabricated citations or biased analysis. In addition, researchers must verify that any third-party AI platform complies with strict data privacy rules. This step ensures that unpublished research data and confidential findings remain fully protected.
Strict requirements for AI disclosures
Transparency remains a central pillar of the updated Elsevier framework. Authors must disclose the use of generative AI by including a formal declaration statement upon manuscript submission. This declaration must state the specific tool used and describe its exact purpose within the preparation process.
While basic spelling and punctuation edits do not require disclosure, substantive structural changes must be declared. Furthermore, if AI tools are used within the actual research methodology, authors must detail this process in the Methods section to ensure scientific reproducibility.
New standards for images and artwork
One of the most notable updates concerns figures, schematics, and observational data. Elsevier now distinguishes between conceptual explanatory diagrams and primary scientific evidence. Authors may use AI to create illustrative flowcharts or decision trees, provided the tool is named in the figure caption.
Conversely, the rules for primary data images are strictly enforced. AI tools must not be used to manipulate, fabricate, or alter observational data. This restriction applies directly to microscopy, histology, radiology scans, and western blots. In addition, general generative AI tools cannot be used to produce graphical abstracts.
Protecting confidentiality in peer review
For peer reviewers and journal editors, confidentiality requirements remain paramount. Reviewers are strictly prohibited from uploading submitted manuscripts into public or open AI models. Doing so risks violating author copyright and data privacy laws.
Reviewers and editors may only use private AI tools that do not retain or re-use uploaded data. Furthermore, these tools should only serve a supportive role, such as improving the readability of decision letters. The critical assessment of scientific merit must remain an entirely human responsibility.
Finally, Elsevier confirmed that its internal editorial tools follow the RELX Responsible AI Principles. These systems assist with reviewer selection and duplicate detection while enforcing continuous bias monitoring.
