AI Disclosure Isn’t the Same Everywhere: What business and social science journals actually require

Most guidance on disclosing generative AI reads as if all journals want the same thing. They don’t. The advice researchers most often encounter tend to focus on publishers like Nature and Elsevier, or guidelines from COPE and ICMJE. That is useful, but for researchers submitting to a business or social science title, those default recommendations can be a poor fit, because the publishers that dominate those fields have their own rules that do not always align with STEM standards.

Consider Emerald, whose portfolio spans management, marketing, information science, and applied social research. Emerald draws a hard line before disclosure even enters the picture: using a generative tool to create, draft, or write new material (the abstract, the introduction, the literature review, the conclusion) is not permitted. Copy-editing an author’s own existing text to improve clarity and grammar is allowed. AI use must be disclosed clearly and transparently within the manuscript, in the Methods and Acknowledgments or another appropriate section, and the author must name the specific tool and its version, and describe what it was used for. Emerald also explicitly states that a language model cannot be credited as an author, and that responsibility for the accuracy of citations and references remains with the human author.

Sage, equally central to the social sciences, frames the question differently. Rather than banning categories of AI use outright, its framework distinguishes between assistive and generative AI. Assistive AI, such as tools that improve language, grammar, and structure in the author’s own writing, does not require disclosure. Generative AI that produces text, references, or images, however, must be disclosed, including the AI model used and the purpose for which it was used. What Sage emphasizes most is verification: when a generative tool is involved, the burden of confirming factual accuracy sits squarely with the author. That extends pointedly to the reference list. Authors are expected to check that every citation is real and correct, to watch for fabrication, and to cite original sources rather than the AI tool itself. Properly disclosed AI use is unlikely to adversely impact a submission; an undisclosed application that produced invented references can.

To see how these frameworks differ from STEM norms, compare them with engineering giants like IEEE. IEEE requires AI-generated content to be disclosed in the acknowledgments but asks for greater granularity: name the system, identify which sections used it, and briefly explain the extent of that use. It even advises authors not to run their reference section through AI editing tools at all, to avoid unwanted changes. Therefore, the landscape isn’t “business versus everyone else.” It runs from Emerald’s prohibition on generating content, through Sage’s disclose-and-verify model, to IEEE’s section-by-section accounting, three genuinely different asks. Enago’s Responsible Use of AI (RUAI) initiative includes a Publisher AI Guidelines Hub that tracks how individual publishers approach AI use, helping authors identify the requirements that apply to their target journal.

Quick reference

The practical takeaway is that “I disclosed my AI use” is not a portable statement. A disclosure written for one journal’s expectations may under- or over-report when evaluated by another journal, and that gap can surface at desk-check. The sensible first move is to write the statement based on the actual policy of the target journal. Tools such as Enago’s AI Disclosure Statement Generator can help authors prepare a structured first draft, but the final disclosure should always be reviewed against the journal’s specific requirements.

In business and the social sciences publication, the publisher defines responsible AI use, not a general rule of thumb. Knowing which rulebook applies is the difference between a smooth submission and an avoidable rejection. The deeper risk in these fields is that AI may quietly introduce a subtly misstated finding or a reference the author never actually read. Those are precisely the problems a generator cannot catch, because they live in the manuscript itself. This is where review by subject-matter experts becomes invaluable. Enago’s Advance Editing pairs authors with editors in their own discipline who can check accuracy, tighten structure, and align the manuscript and its references with the target journal’s requirements before an editor ever sees it.