AI Policies.

COPE Recommended AI Policies

The Committee on Publication Ethics (COPE) has developed a set of fundamental recommendations for the ethical and transparent use of artificial intelligence in scientific publishing. The key policies, based on current COPE guidelines, are presented below.

1. Fundamental Principles
1.1. AI Cannot Be an Author
COPE states unequivocally that AI tools cannot be listed as authors of any scientific manuscript. This is because authorship entails legal and ethical responsibility for the content, as well as the ability to give consent and manage conflicts of interest—qualities that AI does not possess.

1.2. Human Responsibility
Human authors are solely responsible for all submitted content, including the manuscript's originality, accuracy, integrity, and ethical and legal compliance. This responsibility cannot be delegated to an AI tool.

1.3. Mandatory Transparency
Any use of AI beyond routine language assistance (grammar correction, typographical editing, clarity enhancement) must be clearly and thoroughly declared. Transparency is a core pillar of COPE's policy.

2. Policies for Authors
Type of Permitted Use Requirements
Text editing (grammar, style, structure) Yes, with declaration Declare tool and purpose
Literature search Yes, with declaration Ensure it does not generate false citations
Data analysis (ML libraries) Yes, with declaration Describe in the "Methodology" section
Complete content generation No Do not create entire text (introduction, conclusions)
Generation of fictitious data No Do not create synthetic datasets to falsify results
Undeclared use No Mandatory to mention tools in the corresponding section

Additional recommendation: It is suggested to follow structured frameworks such as the one proposed by GAIDeT (Generative AI Declaration Tool) to document in a granular way which tasks were delegated to AI and under what human responsibility.

3. Policies for Peer Reviewers

Type of Permitted Use? Requirements
Linguistic assistance (polishing comments) Yes, with declaration Declare use of AI in the review
Search for Reviewers (automated selection) Yes Based on keywords
Uploading manuscripts to public systems (ChatGPT, etc.) Prohibited Violates confidentiality and copyright
Generating full review with AI No Expert evaluation requires human judgment
Not declaring AI use No Mandatory disclosure

Warning: Editors have reported an increase in AI-generated reviews that appear detailed but contain inaccurate or fabricated criticisms. The fundamental purpose of peer review is to obtain expert human judgment, which AI cannot reliably provide.

4. Policies for Editors

Type of Use Permitted? Requirements
Plagiarism detection (iThenticate) Yes Standard tools
AI-generated text detection Yes Consider probabilistic nature
AI-assisted editorial editing With caution Must not dilute the author's voice or change the meaning
Uploading manuscripts to public systems Prohibited Same as reviewers

Considerations on AI-assisted editorial editing:
- Exists A blurred line exists between assistive use (grammatical corrections, clarity) and generative use (systematic rewriting that alters the author's voice).

- COPE recommends that editors:

1. Establish clear policies defining what constitutes assistive vs. generative use.

2. Evaluate whether changes proposed by AI would be made by a human proofreader.

3. In humanities and social science disciplines, the author's voice is part of the content; this should be considered.

4. Be transparent with authors about the tools used post-acceptance.

5. AI Use Declaration (Suggested Model)
Based on COPE's recommendations, the following format is suggested for the authors' declaration:
> Declaration of AI Tool Use
> During the preparation of this work, the authors used [tool name] for [specific purpose: style correction / language improvement / bibliographic search assistance / data analysis]. After using this tool, the authors reviewed and edited the content accordingly. It was necessary, and they assume full responsibility for the final content of the publication. [The authors declare that no AI tools were used in the preparation of this manuscript / the following tools were used...]
For more complex uses, it is recommended to employ structured disclosure frameworks such as the one proposed in GAIDeT or to follow the guidelines of the Vancouver Standard currently under development.

6. Initiatives in Development

COPE is actively involved in the development of global standards for AI disclosure in research:
Vancouver Standard (2026): COPE has joined ISC, WCRIF, STM, and GYA to develop

In Science and Philosophy, we are aware that “in the vast majority of articles, as AI language models appear to be expanding at an exponential rate and their capabilities tend to develop at an unprecedented pace, [this has led to] significant controversies and dilemmas in terms of how they will change academic writing and knowledge production.”

For this reason, we have divided the policy on Artificial Intelligence into the following statements.

AI is not an author; it is a research assistant that is useful for explaining articles, exploring data, and formatting citations.

Risks: depend on a) biases in the data; b) incorrect, inaccurate, or misleading information of "critical importance," particularly hallucinated references reflected in academic works, in completely fabricated data used for empirical purposes.

This raises concerns about the reliability of the tool in scientific writing. Texts created with AI show a high error rate from a factual perspective, which calls into question their use in certain areas of research, such as bibliometric analysis.

Regarding quality, AI-generated texts tend to be of lower quality and comprehensiveness; however, it is also worth emphasizing that it is increasingly difficult to differentiate between AI-generated texts and human-written texts, particularly in texts that raise concerns about research accuracy and the compromise of quality for the sake of productivity.

Ethics: concerns about the definition of authorship, accountability, and the application of ethical standards in academic publishing. The legal aspect of ethical concerns includes risks such as copyright violations and plagiarism arising from unauthorised information. Verified or incorrectly generated by ChatGPT.

Concerns about “creativity” can no longer be limited to writing ability, but also to using ChatGPT or other large language models (LLMs) to write creatively,” which ultimately questions the very essence of authorship and academic writing and may be a central idea in the era of AI-assisted research.

Academic Innovation and Integration: The use of ChatGPT for research leads to an increase in AI hallucinations, especially in terms of references, which may be alarming given the growing issues surrounding the lack of capacity, or in certain cases, rigor, in peer review.

We recommend the cautious use of AI in scientific literature to preserve publication standards and reliability with rigorous human oversight, i.e., strict adherence to copyright laws and academic regulations when using AI tools to limit potential harm.

It should also be noted that academic integrity, although endangered by the excessive use of ChatGPT in writing, can be protected, as it is easier to detect or highlight data errors or false data or results through ChatGPT.

Self-regulation, critical thinking, and ethical interaction with AI in educational contexts are encouraged; any intention that is unrelated to the AI policy proposed here will be grounds for article rejection.

Reference

Lenvdai GF. ChatGPT in academic writing: A scientometric analysis of the literature published between 2022 and 2023. Journal of Empirical Research on Human Research Ethics. 2025;20(3):131-148. doi: 10.1177/15562646251350203