A journal receives a manuscript that reads clearly, follows academic conventions, and presents a convincing argument. But some of its text may have been generated or revised using AI. Should the editor reject it, ask for disclosure, or examine the research more closely?<\/p>\n
This is becoming an important question in scientific publishing. AI tools can assist with language editing, literature reviews, data analysis, and manuscript preparation. They can also introduce fabricated references, unsupported claims, and questions about who is accountable for the published work.<\/p>\n
AI detection in scientific publishing is moving beyond identifying AI-generated text toward a broader approach to research integrity, transparency, and accountability.<\/strong> For journal editors, reviewers, publishers, and research integrity professionals, the challenge is to distinguish acceptable assistance from practices that compromise the reliability of scientific work.<\/p>\n AI can enter the publishing process at several stages, and each raises different concerns.<\/p>\n Manuscript preparation:<\/strong> Authors may use AI to improve grammar, translate text, organize ideas, or generate new paragraphs. These activities are not equivalent. Language editing may help researchers communicate their findings, while generated content may introduce claims that require careful verification.<\/p>\n Peer review:<\/strong> Reviewers may use AI to summarize manuscripts or help draft reports. However, submitting an unpublished manuscript to an unapproved AI tool can expose confidential research. Reviewers must also ensure that their assessments reflect their own expert judgment.<\/p>\n Editorial screening:<\/strong> Editors may use automated tools to identify potential AI-generated text or other integrity concerns. Yet a detection score cannot establish whether an author violated a journal’s policy, and it cannot verify the scientific validity of a study.<\/p>\n These different uses require different safeguards rather than a single rule for all AI involvement.<\/p>\n The World Association of Medical Editors (WAME) recommends that AI chatbots should not be listed as authors, that authors disclose relevant AI use, and that human authors remain responsible for the accuracy and integrity of their work. Its guidance also addresses the disclosure of AI use by editors and peer reviewers.<\/p>\n Read WAME’s recommendations on generative AI and scholarly publications<\/a>.<\/p>\n Disclosure helps editors understand how AI contributed to a manuscript, but it does not replace verification. An author who discloses using AI to draft text remains responsible for checking its claims, references, and interpretation of results.<\/p>\n Likewise, a manuscript that receives a low AI detection score is not necessarily scientifically sound. Research integrity also depends on accurate data, appropriate methods, valid citations, and honest reporting.<\/p>\n AI detection can help identify text that warrants closer examination, but it should not function as an automatic rejection mechanism.<\/p>\n For example, an editor reviewing a manuscript might encounter a high AI detection score alongside several references that do not support the claims attributed to them. The appropriate response is to verify the references, examine the relevant passages, and follow the journal’s established integrity procedures. The score alone does not prove misconduct.<\/p>\n Conversely, a low score does not establish that a manuscript is original or that its scientific claims are correct.<\/p>\nAI Use Is Changing More Than Manuscript Writing<\/h2>\n
Why Are Disclosure and Accountability Becoming More Important?<\/h2>\n
What Role Should AI Detection Play in Peer Review?<\/h2>\n