Policies on using artificial intelligence adopted by journals in psychiatry and mental health

This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0). European Science Editing / ese Citation Cite this article as: Baminiwatta A, Costa C, Weerasinghe D, Yasir Arafat SM, Lund BD. Policies on using artificial intelligence adopted by journals in psychiatry and mental health. Eur Sci Ed. 2025;51:e165365. https://doi.org/10.3897/ese.2025.e165365 Received: 20 Jul 2025 Accepted: 27 Aug 2025 Published: 27 Oct 2025 Original Article Policies on the use of artificial Policies on the use of artificial intelligence adopted by journals in intelligence adopted by journals in psychiatry and mental healthpsychiatry and mental health Anuradha Baminiwatta1, Chathuranga Costa2, Dinuka Weerasinghe2, S.M. YasirArafat3,4, BradyD. Lund5 1Department of Psychiatry, Faculty of Medicine, University of Kelaniya, Ragama, Sri Lanka [email protected] orcid.org/0000-0002-5495-2029 2Psychiatry Unit, Colombo North Teaching Hospital, Ragama, Sri Lanka 3Department of Psychiatry, Bangladesh Specialized Hospital, Dhaka, Bangladesh 4Department of Public and Community Health, Faculty of Medicine and Health Sciences, Frontier University, Garowe, Somalia orcid.org/0000-0003-0521-5708 5Department of Information Science, University of North Texas, Denton, TX, USA orcid.org/0000-0002-4819-8162 Baminiwatta etal. Policies on using AI adopted by journals X XX https://doi.org/10.3897/ese.2025.e165365 Data Availability Statement The data underpinning the analysis reported in this paper are deposited in the Open Science Framework (https:// doi. org/ 10.17605 /OSF.IO/ BDA4T). Declaration of Interests The authors have no conflict of interest to declare. Funding The authors declared that this study has received no financial support. Compliance with regulations on ethics Ethical approval was not required for this study as it involved analysis of publicly accessible information (journal policies) and did not include human participants, animals, or identifiable personal data. Baminiwatta etal. / https://doi.org/10.3897/ese.2025.e165365 Page 2 / 15 European Science Editing / ese Abstract Background: The integration of artificial intelligence (AI) tools in academic publishing is expanding rapidly, raising concerns about authorship, transparency, and editorial standards. Although organisations such as Committee on Publication Ethics and International Council of Medical Journal Editors have proposed guidelines on the use of AI, the extent to which they have been adopted by journals in psychiatry and mental health remains unclear. Objectives: To examine the adoption and content of AI policies in psychiatry and mental health journals indexed in SCImago and to determine whether higher-quartile journals are more likely to include policies related to AI. Methods: Policies related to AI in the guidelines for authors and reviewers were examined for two groups of journals, all indexed under Psychiatry and Mental Health in SCImago in November-December 2024. The two groups were (1) a stratified random sample of 200 journals (50 per quartile) chosen from a total of 578 journals and (2) 25 top-ranked journals in psychiatry and mental health. Results: Among the first group, 78 (39%) journals included policies related to AI in their guidelines or instructions for authors and reviewers, the number being greater in top-quartile journals (56% in Q1 versus 20% in Q4; χ² = 14, P = .003). Of the 78 journals, 69 (88.5%) disallowed AI tools as named authors, an equal number mandated disclosure of the use of AI, and 58 (74.4%) emphasised author accountability. Peer review policies mostly prohibited AI use (n = 47); AI-assisted copy editing was permitted in 56 journals; and policies on AI-generated images varied. None reported using AI detection tools. Among the top 25 journals, 16 (64%) included policies related to AI; all prohibited authorship to AI and required disclosure; and one reported using AI detection tools. Conclusion: Despite the rising use of AI in publishing, most psychiatry and mental health journals, especially the lower-quartile journals, lack policies on such use. Wider adoption and standardisation of policies related to AI are crucial to ensure research integrity and credibility. Keywords: AI policy, artificial intelligence, large language models, psychiatry, publishing ethics Policies on using AI adopted by journals Baminiwatta etal. / https://doi.org/10.3897/ese.2025.e165365 Page 3 / 15 Introduction Chatbots such as ChatGPT, which use AI (artificial intelligence) and are built on large language models, offer enhanced efficiency and convenience for scholarly work.1,2 Recent surveys highlight the growing popularity of AI chatbots among researchers. An international survey involving more than 2000 medical researchers from 95 countries found that 44.5% used AI chatbots in research, mostly for writing or editing manuscripts, with 62% of the respondents viewing AI chatbots as helpful for general administrative tasks.3 Another survey of more than 2300 researchers representing various regions, disciplines, and career stages showed that 76% had used AI in their research-related activities.4 With the rapid advances and growing use of AI technology internationally, the ethical and practical implications of the use of AI for academic publishing need to be regularly explored. A range of ethical concerns have been raised about the use of AI in scientific publishing, particularly regarding authorship, attribution, copyright, plagiarism, and potential biases.1,2 Researchers may use AI to increase their rate of research output and, without adequate human oversight, the proliferation of such content may compromise the integrity and validity of scientific publications.5,6 Groups such as the Committee of Publication Ethics (COPE), International Committee of Medical Journal Editors (ICMJE), and World Association of Medical Editors have proposed guidelines regarding the use of AI within the publishing process.7-9 Perkins and Roe conducted a thematic analysis of publisher guidelines on AI use and identified six themes: confining authorship only to humans (and not to any other entities including AI tools), author accountability, disclosure and transparency, research integrity, fluid policies, and constraints and exclusions.10 Given the potential for unethical use of AI in scientific publishing, all scientific journals and publishers need to provide clear guidelines to authors and reviewers regarding the use of AI.11 However, by the end of 2023, only 56 out of 162 (34.6%) publishers affiliated to the International Association of Scientific, Technical and Medical Publishers had a publicly available AI policy.12 According to Ganjavi etal., among the top 100 publishers in terms of size, only 24 provided an AI guideline, of which 15 were among the top 25 publishers; in contrast, among the top 100 journals, 87 provided guidance on AI.13 Similarly, Lund etal14 reported that among the top 300 academic journals, more than half had an AI policy. Only a few studies have focused on specific disciplines. A study of ophthalmology journals indexed in PubMed found that among the 84 journals in the field, 53 (63.1%) had AI policies, and that journals with AI policies had higher metrics including the impact factor and SCImago Journal Rank (SJR).15 Although journals tend to adopt the publishing policies of their publishers, substantial variability was observed in AI policies even among journals published by the same publisher.13 The use of AI has been increasing in psychiatric research in recent years.16 The implications of AI for academic publishing specifically within the field of psychiatry have also been discussed, and largely reflect the general concerns raised in academia in general.17-19 Mental health researchers, editors, and reviewers of journals in psychiatry need to keep themselves abreast of advancements in AI and their implications for scientific publishing. Unless psychiatry journals provide clear guidelines on the use of AI, there is a risk of unethical and improper use that could compromise the scientific validity and ethical integrity of publications in this field. This study therefore aims to (1) examine the current adoption of AI policies in psychiatry Policies on using AI adopted by journals Baminiwatta etal. / https://doi.org/10.3897/ese.2025.e165365 Page 4 / 15 and mental health journals across all four SCImago quartiles, (2) summarise the content of these policies, and (3) assess whether higher-quartile journals are more likely to include AI policies. Methods Selection of journals All journals indexed under Psychiatry and Mental Health in SCImago (a public portal reporting journaland country-level indicators based on the Scopus® database) were reviewed, as recorded in its 2023 list, on 3 November 2024 and then divided into two categories: (1) a stratified random sample of 200 journals (50 per quartile) out of the 578 journals listed and (2) top 25 journals in psychiatry and mental health. SCImago ranks journals using SJR, which takes into account both citation counts and the prestige of citing journals. Quartiles are assigned by dividing the ordered list into four groups: Q1 (top 25%), Q2, Q3, and Q4 (bottom 25%). The random sample of 200 journals was to estimate the proportion of journals adopting AI policies with a margin of error of ±5.5% at a 95% confidence level. For the random selection, random numbers were generated from the website https:// www.rand om.org. Ten of the top 25 journals also featured in the 50 journals selected randomly from Q1. The study methodology was pre-registered on the Open Science Framework.20 The journal selection process is summarised in Figure 1. Data extraction In November-December 2024, two reviewers independently extracted the details of policies related to the use of AI from the guidelines for authors and reviewers available on the official website of each journal. Disagreements between the two reviewers were resolved with inputs from the principal investigator. The information extracted for each selected journal is summarised in Table 1. Ethics Ethical approval was not required for this study as it involved analysis of publicly accessible information (journal policies) and did not include human participants, animals, or identifiable personal data. Data synthesis and analysis Extracted policy data from the journals were qualitatively synthesised. We used chi-square tests to ascertain whether some characteristics of a given journal (quartile, affiliation with a major publisher, origin from the United States or the UK) made it more likely or less likely to include AI policies. Results Characteristics of selected journals Characteristics of the 200 randomly selected journals and those of the top 25 are summarised in Table 2. Among the 200 journals, the leading publishers were Elsevier, Springer Nature, Wiley, Sage, Routledge, Cambridge University Press, Lippincott Williams & Wilkins, Taylor & Francis, American Psychological Association, and Emerald Publishing; these publishers accounted for 60% of the journals. Most journals were based in the United States (n = 62) or the United Kingdom (n = 61). The combined representation of the two countries was 82% in Q1, 72% in Q2, 68% in Q3, and 24% in Q4 journals. Among the top 25 journals, the leading publishers were Elsevier, Wiley, Springer Nature, and Taylor & Francis. Except for one journal from Switzerland, all the others were from either the United States or the United Kingdom. Policies on using AI adopted by journals Baminiwatta etal. / https://doi.org/10.3897/ese.2025.e165365 Page 5 / 15 Adoption of policies on the use of artificial intelligence Of the 200 journals, 78 (39%; 95% CI: 33.5%– 44.5%) included policies on the use of AI within the guidelines for authors or reviewers; among the top 25, 16 (64%) had such policies. Higher-quartile journals were more likely to have AI policies (χ2 = 14, P = .003): 56% of Q1, 42% of Q2, 36% of Q3, and 20% of Q4 journals had such policies. Among the 120 journals published by the top ten publishers, 59 (49.2%) included AI policies, compared to 19 (23.8%) of the 80 journals published by other publishers (χ2 = 13, P < .001). Journals based in the United States or the UK were also more likely to include AI policies (χ2 = 7.2, P = .007): 46.3% of the 123 US/UK journals versus 27.3% of the journals published from elsewhere. No artificial intelligence as author Of the 78 journals that did include policies on the use of AI, 69 (88.5%) explicitly prohibited AI as a named author, typically citing the inability of AI tools to assume responsibility for submitted work (Table 3). The remaining nine did not state their position clearly. Among the top 25 journals, all 16 with AI policies explicitly prohibited AI authorship (Table 4). Figure1. Identification and selection of journals for artificial intelligence policy review. Policies on using AI adopted by journals Baminiwatta etal. / https://doi.org/10.3897/ese.2025.e165365 Page 6 / 15 Author accountability Of the 78 journals, 58 (74.4%) emphasised that authors are fully responsible and accountable for all AI-assisted content; 8 journals were unclear on this matter; and the remaining 12 journals did not directly comment on the authors’ accountability regarding AI-generated content. Among the top 25 journals, 13 out of 16 (81.2%) with AI policies mentioned the authors’ accountability for AI-generated content. Transparency and disclosure Of the 78 journals, 69 (88.5%) required transparency and insisted on disclosure of the fact if authors had used AI. However, the journals differed in their recommendations on the specific section of the manuscript that should feature the disclosure: of the 69, 31 recommended the methods section; 16 stipulated a separate section; 16 more recommended either the methods section or the acknowledgements; 5 recommended the acknowledgements; and one journal recommended the cover letter. Of the 16 that offered a choice, the choice was guided by the purpose for which AI had been used: if for help in writing, the acknowledgements were considered the appropriate place; if for help in formal research design and methods including data collection, analysis, or figure generation, it was the methods section. Some journals required details of the specific AI tool used (name, version, and manufacturer) and how it was applied. Four journals recommended an additional disclosure of the use of AI in the cover letter at initial submission. Five journals, all published by the American Psychological Association, required Table 1. Elements related to artificial intelligence policy extracted from the guidelines to authors and reviewers given by the journals chosen for the study Aspect of AI policy Description AI policy inclusion Is policy on AI included in the guidelines? Source guideline Do the guidelines cite any other guideline on the use of AI? If so, which of these guidelines? (For example, ICMJE, COPE, or WAME) Policy on AI tool as an author Is authorship for AI tools allowed or prohibited? Author accountability Do the guidelines state that authors should be accountable for any AI-generated content in the manuscript? Transparency and disclosure Do the author guidelines require disclosure of the use of AI? If so, in which section or part of the manuscript should this disclosure be placed? Research integrity Do the author guidelines flag the potential risks of AI use on the quality and integrity of the work? Are authors required to discuss these issues within the manuscript? Adaptability of policies Do the guidelines state that the AI policies may be updated or revised in the future? If so, are updates to be expected at regular intervals? Constraints and exclusions Does the policy explicitly prohibit the use of AI? AI-generated images What is the journal’s policy on the submission of images (figures, photographs, etc.) created by generative AI? AI-assisted copy editing What is the journal’s policy on AI-assisted improvements to humangenerated texts for readability, style, grammar, and spelling? Peer review policy Does the journal allow or prohibit peer reviewers to use AI? AI detection tools Does the journal use any AI detection tools to screen submissions for AI-generated content? AI: artificial intelligence; COPE: Committee of Publication Ethics; ICMJE: International Committee of Medical Journal Editors; WAME: World Association of Medical Editors. Policies on using AI adopted by journals Baminiwatta etal. / https://doi.org/10.3897/ese.2025.e165365 Page 7 / 15 that the AI outputs be uploaded as supplementary material. Among the top 25 journals, all 16 that included AI policies required transparency and disclosure regarding the use of AI. Of the 16, the recommended sections for disclosure were the methods section or a separate section (5 journals each), a choice between the methods section and the acknowledgements (3 journals), acknowledgements only (2 journals), and disclosure in “both the relevant text of the submitted work and the cover letter or email accompanying the submission” (1 journal). Integrity concerns Only 14 (17.9%) of the 78 journals addressed the potential impact of the use of AI on research integrity and ethics, and did so with Table 2. Characteristics of 200 journals selected randomly from 578 journals listed under psychiatry and mental health and the top 25 journals in that domain Characteristic Randomly selected journals Top 25 journals Major publishers Elsevier 20 6 Springer Nature 19 3 Wiley 17 5 SAGE 13 – Routledge 13 – Cambridge University Press 10 1 Lippincott Williams & Wilkins 8 – Taylor & Francis 7 2 American Psychological Association 7 – Emerald Publishing 6– Other publishers 80 8 Country of origin United States 62 12 United Kingdom 61 12 The Netherlands 10 – Switzerland 8 1 Germany 8 – Spain 5– Brazil 4– Canada 4– India 4– Iran 4– Other countries 30 – Table 3. Policy elements covered by psychiatry and mental health journals that include policies on artificial intelligence (among the 200 randomly selected journals) Policy element Journals (n = 78) Number Proportion (%) AI tool as a named author 69 88.5 Author accountability 58 74.4 Transparency and disclosure 69 88.5 Integrity concerns 14 17.9 Adaptability of policies 27 34.6 Prohibition on any use of AI 00.0 Peer review 49 62.8 AI-generated images 55 70.5 AI-assisted copy editing 56 71.8 AI detection tools 00.0 AI: artificial intelligence. Table 4. Policy elements covered by psychiatry and mental health journals that include policies on artificial intelligence (among the top 25 journals) Policy element Journals (n = 16) Number Proportion (%) AI tool as named author 16 100.0 Author accountability 13 81.2 Transparency and disclosure 16 100.0 Integrity concerns 425.0 Adaptability of policies 7 43.8 Prohibition of any use of AI 00.0 Peer review 9 56.2 AI-generated images 10 62.5 AI-assisted copy editing 850.0 AI detection tools 1 6.2 AI: artificial intelligence. Policies on using AI adopted by journals Baminiwatta etal. / https://doi.org/10.3897/ese.2025.e165365 Page 8 / 15 little consistency among them. For instance, some journals published by Mary Ann Liebert publishers mentioned that “potential biases and limitations of the outcomes of AI use should be discussed by the authors when presenting their results” and some journals published by Elsevier included the following cautionary note: “AI can generate authoritative sounding output that can be incorrect, incomplete, or biased.” Among the top 25 journals, 4 out of the 16 (25%) with AI policies addressed integrity concerns. Adaptability of policies Roughly a third (34.6%, or 27 out of 78) of the journals noted that AI policies would be updated periodically as AI is a rapidly advancing field. For example, two journals, published by Mary Ann Liebert publishers, stated that “the policies below will be reviewed and updated as technologies, best practices and ethical considerations in AI evolve.” However, the frequency of reviewing and updating policies was not stated in any of the journals. Among the top 25 journals, 7 of the 16 (43.8%) with AI policies commented on the adaptability of policies, with one journal specifying the frequency of updates (every 6 months or as necessary). Policy on artificial intelligence–generated images Policies on AI-generated images varied across journals. Roughly half of them (48.7%, or 38 out of 78) prohibited or restricted the use of AI-generated images, although 29 of them outlined specific exceptions under which such images could be permitted; 17 adopted a more flexible stance, allowing AI-generated images but mandating disclosure; and 23 provided no information on AI-generated images within their AI policies. The nature of exceptions varied with the publisher. Springer Nature journals permitted AI-generated images in some cases, such as artwork obtained from agencies with contractual agreements with the publisher and created through legally acceptable means or images referenced in content specifically discussing AI, subject to a case-by-case editorial review. Elsevier journals noted that exceptions may be made when AI is integral to the research design or methodology, for instance in AI-assisted imaging techniques used for generating or interpreting research data, such as in biomedical imaging. They also stated that AI-generated cover art could be allowed, provided prior approval is obtained from both the editor and the publisher. Emerald allowed the use of generative AI images for illustrative purposes within scholarly critique and discussion. Among the top 25 journals, 10 of the 16 (62.5%) with AI policies included a policy on AI-generated images. Eight journals prohibited or restricted the use of AI images, although six of them outlined specific exceptions. The remaining two journals adopted a more flexible stance, allowing AI-generated images but requiring disclosure. Artificial intelligence–assisted copy editing Artificial intelligence–assisted copy editing was allowed by 56 of the 78 journals (69.2%), of which 15 required that such use be disclosed, whereas the rest did not mandate a disclosure; 22 journals did not explicitly mention a policy on this. Among the top 25 journals, 50% Policies on using AI adopted by journals Baminiwatta etal. / https://doi.org/10.3897/ese.2025.e165365 Page 9 / 15 of journals with AI policies included a policy on AI-assisted copy editing. Peer review The use of AI in peer review was either prohibited or restricted by 47 of the 78 journals (60.3%). Most of them advised reviewers not to upload manuscripts into generative AI tools. Two journals were less restrictive but required disclosure of the use of AI. Some journals, particularly those published by Springer Nature or Elsevier, mentioned that they are currently exploring safe and compliant AI tools for reviewers, and may update policies accordingly. Among the top 25 journals, 9 of the 16 (56.25%) journals with AI policies addressed the use of AI in peer review. Tools for detecting the use of artificial intelligence None of the 200 randomly selected journals mentioned whether they use AI detection tools. Only one of the top 25 journals (published by BMJ Publishing Group) reported using AI detection tools. Discussion Our study revealed that only 39% of the journals in the field of psychiatry and mental health that formed the basis of our data have adopted formal policies regulating the use of AI. This small proportion raises major concerns about the integrity and accountability of scientific papers published in this field since the popularisation of AI chatbots. It is noteworthy that this percentage is remarkably lower than that reported for journals in some other disciplines such as ophthalmology (63.1%)15 and dentistry (70.9%).21 In parallel with the present study, Tran etal. reviewed the top 100 psychiatry and mental health journals on SCImago and found that 91% addressed the use of AI in their instructions to authors; 22% allowed AI-generated content; and 13% permitted AI-generated images.22 Our lower estimate (39% versus 91%) likely reflects that our sample spans all quartiles, rather than only top-ranked journals. As journal policies keep evolving, repeating this study in the future could chart the progress made in this respect. We found that higher-ranking journals were more likely to employ AI policies, with 56% of the Q1 journals versus only 20% of the Q4 journals including a policy on AI. This finding aligns with observations in other disciplines.15,23,24 Affiliation with a major publisher, more frequent policy updates, and stronger editorial awareness on AI policies are some possible reasons for the higher-impact journals having AI policies. More importantly, these observations call for urgent updates to the author and reviewer guidelines in psychiatry and mental health journals, especially by the lower-quartile journals. We also observed that fewer journals originating from outside the United States and the UK included AI policies. Journals from lowand middle-income countries (LMICs) may face challenges in adopting AI policies, potentially because of limited familiarity with AI technologies among members of their editorial boards. Greater standardisation of AI policies among academic journals will ensure consistency and fairness in approaching the use of AI in scholarship. However, given that the current guidelines from different committees, publishers, and journals show inconsistencies, it would be helpful for future researchers, editors, and reviewers to be able to refer to a consensus guideline on the use of AI. Efforts to this end are already in progress; for example, the