Generative-AI and AI-assisted tools cannot be authors, reviewers or editors. Accountability for every submitted, reviewed and published word rests with identifiable people. This policy sets out separate obligations for authors, for reviewers and for editors, because the duties of each role differ.
1. Scope and definitions
- Generative AI — a tool that produces new text, images, code or data on request: a large language model, a chatbot, or an image generator.
- AI-assisted tool — a tool that transforms material the user supplies: grammar correction, paraphrasing, translation, transcription, or reference formatting.
- Legitimate computational method — an established analytical technique reported as part of the research itself: topic modelling, sentiment or content analysis, machine-learning classification, network analysis, or statistical modelling. These are not covered by the disclosure rules in section 3. They must be reported in the methods in enough detail to be reproduced, with the model, parameters and validation stated, and must be clearly distinguished from generative content.
2. Artificial intelligence cannot be an author
An AI tool cannot consent to submission, approve the final version, or take responsibility for the work, and therefore cannot meet the authorship criteria. It must not be listed as an author or co-author, nor named in the contributorship statement as if it were a person. See Authorship.
3. Obligations of authors: disclosure
Authors must disclose any material use of generative AI or AI-assisted tools in producing the manuscript. Disclosure is a condition of consideration; undisclosed material use is treated as a breach of publication ethics.
What must be disclosed: the name and version of the tool and the approximate date of use; the purpose (drafting a section, summarising literature, paraphrasing, translating, coding qualitative data, generating code); the stage (conception, analysis, drafting, revision, language preparation); and the extent, so that a reader can judge how much of the text or analysis it touched.
Where to disclose it: use that affects the research itself is described in the methods; use confined to language or presentation is described in the acknowledgements. In both cases the corresponding author completes the AI declaration in the cover letter.
What need not be disclosed: routine spelling correction, grammar checking and reference-manager formatting, provided they do not change scholarly content, meaning, data or argument. Where there is doubt, disclose.
Prohibited uses by authors:
- generating, altering or augmenting data — including interview transcripts, survey responses, field notes, quotations attributed to participants, or statistical results — or presenting generated material as an authentic research observation;
- entering identifiable or confidential research material into a public AI service: interview recordings or transcripts, personal data of respondents, data obtained under a confidentiality undertaking, or material covered by an ethics approval that does not permit third-party transfer;
- generating references without reading and verifying each one against the actual source; a reference that does not exist, or that does not support the statement it is cited for, is treated as a fabricated citation;
- generating substantial passages and presenting them as the authors' own original composition without disclosure;
- using a tool to evade similarity or integrity screening, or to disguise text reuse.
Accountability. Authors remain fully responsible for everything in the manuscript, including any part produced with assistance. Every factual claim, figure, quotation and reference must be checked by a human author against the primary source. “The tool produced it” is not a defence, and an error traceable to an undisclosed tool is handled as author error under the Publication Ethics policy.
4. Obligations of reviewers: prohibition
A manuscript under review is a confidential document. Uploading it, or any part of it, to a generative-AI service transfers it to a third party and breaches that confidence, whatever the service's own terms say.
- Reviewers must not upload a manuscript, abstract, table, dataset, transcript, supplementary file, review report or editorial correspondence to any generative-AI or AI-assisted service, public or commercial.
- Reviewers must not use generative AI to write, draft, summarise or compose any part of a review report. The assessment must be the reviewer's own, in the reviewer's own words; see the Peer Review Process for what a report must contain.
- Reviewers must not use AI to attempt to identify an anonymised author, or to search for an unpublished manuscript online.
- A reviewer unable to complete the assessment without such assistance must decline the invitation instead.
- A reviewer who has inadvertently entered confidential material into such a service must inform the editor immediately, so that the authors can be notified and the consequences assessed.
- Routine spelling or grammar checking of the reviewer's own finished comments is permitted, provided no manuscript content is included.
5. Obligations of editors: limits on use
- An editorial decision — desk rejection, choice of reviewers, acceptance, revision or rejection — is made by a named human editor, and the Editor-in-Chief retains responsibility for the final decision. It must never be produced by, delegated to, or substituted by an automated tool, and no tool is given authority to accept or decline a manuscript.
- Editors must not upload manuscripts, review reports, author correspondence or personal data to generative-AI services. The confidentiality owed to authors and reviewers is identical to that owed by reviewers.
- Editors must not use AI to generate decision letters, reviewer comments, or any text presented to an author as a human editorial judgement.
- AI or automated assistance may be used only for tasks involving no confidential manuscript content and no judgement on merit — drafting general correspondence templates, administrative scheduling, or searching public literature to identify candidate reviewers, whose eligibility the editor then verifies against the criteria in the Peer Review Process.
- Where any tool has contributed to editorial or production work on an article, the editor remains accountable for the outcome and the use is recorded in the submission record.
6. Images, figures and data integrity
AI-generated or AI-altered images, figures, maps or data must not be presented as authentic research material. Adjustments to an image must be applied to the whole image, must not obscure or create features, and must be described. Where a generated or illustrative image is published — a schematic or a graphical abstract — the legend must state that it is generated and name the tool. Original, unprocessed files must be retained by the authors and supplied to the journal on request.
7. Translation and language editing
OAIJSS publishes in English and welcomes submissions from authors writing in a second language. Using an AI-assisted tool to translate or improve the language of a manuscript is permitted and is disclosed in the acknowledgements. The authors are responsible for confirming that the translated or edited text still states exactly what they mean — particularly in the methods, the findings and the limitations, and in quotations from participants, where a change of wording can change the scholarly claim or misrepresent what a respondent said.
8. Tools used by the journal
OAIJSS uses ORCA for similarity screening during editorial assessment, as described in the Plagiarism Policy, and EIGO to support copyediting and language proofreading, as described on the Submission Process page.
In every case the output of a tool is decision support only. No tool determines the outcome of a submission: a similarity score, a language report or any automated flag is read and judged by an editor. Any version edited with such support is verified by the editorial team and returned to the corresponding author for approval before the galley proof, and the author remains the author of record for the published text.
9. AI-detection output is not proof
Tools that claim to detect AI-generated text are unreliable, and are known to misclassify work written by authors using English as a second language — a material concern for a journal whose contributors write mainly in a second language. OAIJSS therefore does not reject, retract or accuse on the basis of an AI-detection score. Where a genuine concern arises, the journal examines the manuscript itself: the coherence of the argument, the verifiability of the references, the availability of the underlying data, and the authors' account of how the work was produced. Authors are given the opportunity to respond before any action is taken.
10. Breaches of this policy
Undisclosed material use, fabricated data or references, breach of reviewer or editorial confidentiality, or presenting generated material as authentic is treated as misconduct. Depending on the evidence and the stage, the journal may request clarification, require revision and disclosure, decline the manuscript, publish a correction, publish an expression of concern, retract the article, notify the authors' institution, or decline future submissions from the responsible parties. The procedure, including the applicable COPE flowchart, is set out in the Publication Ethics policy.
11. Review of this policy
This policy follows the COPE position on authorship and AI, the ICMJE Recommendations, and the requirements of Indonesian national journal accreditation. Because these tools change quickly, OAIJSS reviews this policy at least once a year and whenever a material change in practice requires it. Questions about a specific intended use should be sent to indonesiajournalsocialsciences@gmail.com before submission.