An international initiative is currently under way to come to an agreement on a global standard for AI disclosure in research. We are all in some way or another affected by the use of AI in research, either in doing research or in using research outputs. It is important for the academic community to take note of this discussion and to contribute to this initiative, where appropriate.
Despite some convergence on AI disclosure in scholarly publications developing during the past twelve months, the requirements for disclosure are still very different among publishing houses and journals. This causes considerable uncertainty for researchers and reviewers, as well as for readers of such publications.
At the end of 2025, several international science bodies involved in science and guidance to editors of scholarly publications took the initiative to work towards a global reporting standard for AI disclosure in research. Central to this initiative are the following: the International Science Council (ISC), the Committee on Publication Ethics (COPE), the Association of Scientific, Technical and Medical Publishers (STM), the World Conference on Research Integrity (WCRI) and the Global Young Academy (GYA). The ISC has a membership that brings together around 250 scientific organisations around the world, from international scientific associations to national science academies and research councils.
The initiative aims to develop a broadly supported reporting standard that can be used across research disciplines and publication cultures. A Task Team for this global, multi-stakeholder effort came into being and a plan was developed to work across three rounds of consultations and interactions with the user community of scholars and scholarly associations during 2026, leading to a final and refined outcome early in 2027. Round One has been completed, and Round 2 is currently open. (For details on the process, see https://council.science/our-work/ai-disclosure-in-research/.)
The aim of this initiative might sound easily reachable. However, the issues around AI disclosure are complex and multi-faceted. In the case of plagiarism in the pre-GenAI period, it was relatively easy to identify where content from other publications was used without disclosure, and excellent tools were developed to identify such misconduct. In the case of GenAI, new text is generated with every search/chat/action and the conventional plagiarism detectors are not sufficient any longer.
The key questions regarding AI disclosure are about the when, the how and the where of AI disclosure. ‘When?’ is about the question: ‘which use of AI during the research process should be disclosed?’ Evidently, disclosure should be proportionate to the extent of use made of AI, but it also relates in different ways to various tasks undertaken during the research process.
During a special Focus Track of the WCRI in May 2026 in Vancouver, participants were provided with a draft position on AI disclosure and with concrete cases, and they were given the opportunity to indicate which forms of disclosure (or not) would be appropriate to each case. (For the slides, see https://council.science/wp-content/uploads/2026/07/WCRI2026_FocusTrack_2B.pdf.)
In Round Two, “ISC Members and the broader research community” are invited to take part in a survey to work towards a shared, global understanding of AI disclosure by 16 October 2026. This engagement is essentially about the disclosure thresholds and taxonomy. A background document is provided to inform new participants to the discussion on the issues involved. This is essential reading before taking the survey. (Links to the background document and the survey are available at the first link provided above.)
Continued work through Round Two is based on a preliminary statement developed in Round One, namely:
“AI use should be disclosed if it substantively shapes or replaces human judgement, or has another substantive influence on research processes or its reporting. To decide whether an influence is ‘substantive’, researchers should ask: would a disciplinary peer – whether a reviewer or a reader – reasonably need to know about the use of AI to understand how the research was conducted, or to assess the reliability or validity of the published work?”
Detailed forms of AI disclosure are in the first place relevant to science/scholarly publication and in most cases also to funding applications or reports. However, some form of disclosure might also be required for other genres of writing resulting from research, such as policy advice, consultation work and textbooks.
What becomes increasingly clear, is that AI/GenAI affects all dimensions and activities in the academic domain. How it should be disclosed, is only slowly emerging after three and a half years of GenAI use.
Walter Claassen
Acting SFA Lead: Digital Transformation
