AI Tools Usage Policy

Artificial intelligence has become part of how research actually gets done. Authors who submit to JCDB use it to tidy up their English, to debug a stubborn script, to redraw a figure at the last minute, or to track down a paper they only half remember. We see no reason to pretend otherwise, and none to treat these tools as suspect in themselves. What concerns us is something older than any of them: that a reader can trust what the manuscript reports, that the people named on it will stand behind every sentence, and that no claim goes further than the authors can defend.

This document sets out how we expect authors to work with AI tools, when they should tell us about it, and what happens when those expectations are not met. It follows the Core Practices of the Committee on Publication Ethics (COPE) and COPE’s ongoing discussion of artificial intelligence in scholarly publishing.

What we mean by AI tools

We use the term broadly: any software or platform that draws on machine learning, natural language processing, deep learning, or generative models to write, edit, translate, summarize, analyze, visualize, or program. In practice that means large language models such as ChatGPT, Gemini, and Claude; writing assistants such as Grammarly, DeepL Write, and QuillBot; coding assistants such as GitHub Copilot; and the widening range of tools for statistical work, data visualization, literature discovery, reference management, and figure generation.

New tools appear faster than any policy can name them, so treat these examples as illustrations rather than a closed list. The test is simpler than the catalogue: if a tool did a meaningful share of the writing, the analysis, or the making, this policy applies to it.

Where AI can legitimately help

Most uses of AI in manuscript preparation are unremarkable, and we welcome them. Correcting grammar, spelling, and punctuation falls squarely in that category, as does rephrasing a paragraph that reads awkwardly in a second language. Translation is acceptable when a competent human checks the result, since machine translation still misplaces technical terminology in ways that are invisible to the tool but obvious to a specialist. Formatting references, searching the literature, and generating candidate keywords are similarly routine.

On the technical side, authors may use AI assistants to write, debug, or refactor code, to support statistical analysis, and to produce charts, diagrams, and illustrations, provided the underlying data are real, the sources are properly licensed, and the output has been checked by someone who understands what it is meant to show. The condition never changes: the authors have read the result, judged it, corrected it where necessary, and are prepared to answer for it.

Where the line is

Certain uses are never acceptable, and the reason is always the same, they break the link between what the manuscript says and what the authors actually know.

• Fabricating, falsifying, or manipulating data, results, images, or any other form of evidence.
• Presenting references that do not exist. Language models invent citations that look entirely plausible, complete with volume numbers and a well-formed DOI, and every one of them must be verified against the source.
• Reproducing copyrighted or third-party material without permission, whether the tool surfaced it verbatim or paraphrased it closely enough to amount to the same thing.
• Delegating the scientific work itself, the interpretation of results, the judgment about what they mean, the conclusions drawn from them to a model.
• Submitting AI-generated text, code, or figures that no author has genuinely read and revised.
• Concealing AI use in cases where this policy asks for disclosure.

What authors remain responsible for

Responsibility does not transfer to a tool. Everything submitted to JCDB belongs to its authors, including the parts a model helped produce.

In concrete terms, we expect authors to verify AI-assisted content against the underlying evidence, to read and revise generated text, code, figures, and analyses rather than pass them through untouched, and to confirm that every reference is real and correctly cited. Errors, biases, and confidently stated falsehoods are common enough in current systems that catching them is part of the work, not an optional extra. Authors should also think about what they send to third-party services: unpublished results, participant data, and other confidential material may be retained or used for training, and once submitted to a commercial system they are difficult to recall.

Authorship

An AI system cannot be an author or a co-author of a JCDB article. Authorship rests on two things a model cannot supply: a substantial intellectual contribution, and the capacity to take public responsibility for the work when it is questioned. Where a tool’s contribution has been significant enough to be worth recording, acknowledge it in the acknowledgments or in the methods, but it does not belong in the author list.

Disclosing the use of AI

We ask for transparency whenever an AI tool has contributed beyond ordinary proofreading. In practice that means anything more than sentence-level correction of grammar, spelling, or wording: text that reaches the manuscript in substantially the form the tool produced it, analysis or code the tool wrote, figures it generated, or literature it selected. A disclosure need not be long; it only has to let a reader understand what was used and how far it went. Four details are usually enough:

• The name of the tool.
• The version or model, where this is known.
• What it was used for.
• How far its contribution extended.

Authors may adapt the following statement:

Declaration of AI Tool Usage: During the preparation of this manuscript, the authors used [AI tool name and version] for [e.g., language improvement, code assistance, data visualization, or statistical support]. The authors reviewed and revised all AI-assisted content and accept full responsibility for the accuracy, integrity, and originality of the manuscript.

Where the statement belongs depends on what the tool did. If it contributed to analysis, coding, visualization, or any part of the research process, the methods section is the right place, because that is where a reader goes to judge how the results were produced. If it helped only with language or formatting, the acknowledgments will do. A manuscript with more extensive AI involvement may warrant a short section of its own headed “Declaration of AI Tool Usage.” Routine spell-checking and grammar correction with common writing software do not normally need to be declared.

How editors and reviewers handle disclosed use

Editors and reviewers read a disclosure as part of the ordinary methodological and ethical assessment of a submission, not as a mark against it. Where AI use appears inappropriate or seems to have gone undeclared, the editorial office may ask the authors to explain, request revisions or further information, refer the case for a fuller ethical review, or in serious cases decline the manuscript.

JCDB may use screening tools to support this process, but no decision will ever rest on the output of automated AI-detection software alone. These detectors are unreliable in both directions, and they are particularly prone to flagging the careful, conventional prose of authors writing in a second language. A person makes the decision.

When the policy is not followed

The response is proportionate to what has happened. A minor or ambiguous case usually ends with a request for clarification or a correction. Serious cases may lead to rejection, or to retraction if the article has already been published. Where there is evidence of deliberate misconduct we may inform the authors’ institution, and repeated or severe violations may result in restrictions on future submissions.

Appeals

Authors who disagree with an editorial decision concerning AI use may appeal in writing to the Editor-in-Chief. The appeal should set out the grounds for disagreement and include any supporting evidence. It will be considered under the journal’s standard procedures for ethical review.

Our own use of AI

The same standards apply to us. JCDB does not use AI tools to make editorial or peer-review decisions on their own, and we do not intend to. Should AI be introduced into any part of our editorial workflow in future, it will operate under human oversight, and the final judgment will remain with the editorial team.

Revisions to this policy

This is not a settled field, and the policy will change as the technology and the norms around it change. Authors should consult the current version before submitting, and are welcome to contact the editorial office whenever a situation is not clearly covered here. We would much rather answer a question in advance than investigate a problem afterwards.

Ethical framework

This policy draws on the guidance of the Committee on Publication Ethics, in particular the COPE Core Practices, the COPE Discussion Document on Artificial Intelligence, the COPE guidance on authorship and contributorship, and the COPE retraction guidelines. All submissions to the Journal of Computers and Digital Business are expected to meet these standards.