AI policy

This policy details how and when Artificial Intelligence tools can be used by authors of Edinburgh University Press (EUP) books and journals for the purposes of developing a manuscript.

We encourage all authors, editors and reviewers to read this policy in full. Please note that the policy applies to the full planning, writing and publication process. 

As a brief overview, our general positions are:

• All use of AI must be fully and clearly disclosed (see Disclosure and transparency) and the author is responsible for anything that comes from the use of AI, including accuracy, citations and ethical compliance.
• All outputs from AI should be treated as a draft and must be thoroughly checked and verified by a human before they are incorporated into a final manuscript.
• AI can only be used in a responsible, ethical way that does not compromise research integrity (See Use of Assistive AI).
  Whole, unpublished manuscripts (including full books, chapters and journal articles) must not be uploaded to any AI chatbot for any purposes.
• Authors need to assume that anything inputted into an AI chatbot will be used to help train it and could therefore potentially breach author copyright or confidentiality. Authors should be especially wary of free AI tools.
• EUP (or the journal/series editor) can reject a manuscript if we believe that the author has not followed our AI policy.

For a full summary of our positions on use of AI, please see the crib sheet at the end of the policy.

Jump straight to:

• Overall requirements
• Disclosure and transparency
• Use of Assistive AI
• Idea generation
• Peer review
• Citations
• Translation
• Images
• Promotion and book reviews
• Crib sheet: Use of AI

Downloads:

• Download the Edinburgh University Press AI Policy (.pdf)
• Download the Edinburgh University Press AI Disclosure Form (.docx)

Overall requirements

AI can only be used in a responsible way that does not compromise research integrity and ethics. Authors must ensure their use of AI is conducted within transparent scholarly practice, mitigating bias, avoiding plagiarism and preserving confidentiality. 

Edinburgh University Press’s overall requirements for authors’ use of AI are as follows:

Confidentiality & IP protection

• You should treat the information given to any external AI as if you were posting it on a public site (e.g., a social network or a public blog). To prevent violations of the author’s intellectual property (IP), confidentiality and data privacy, including personally identifiable information or AI developers claiming ownership of inputted content or expose sensitive data:
  • Unpublished manuscripts (including full books, chapters and journal articles) must not be uploaded to any AI tools for any purposes.
  • AI tools must not be used to generate the abstract, keywords or any summary of the work.
  • Confidential or personal data must not be uploaded to any AI tools.

Transparency and disclosure

• All use of AI must be fully and clearly disclosed (via an AI Disclosure Form). Authors must never claim AI-generated outputs as their own work without acknowledgement. Never claim AI output as your own work.
• Authors retain legal and scholarly responsibility for AI output, including accuracy, citations and ethical compliance.

Human verification

• AI does not fulfil the requirements for authorship, so the author will be liable for any AI outputs in their work.
• All outputs from AI must be thoroughly checked and verified for accuracy, bias and source attribution by a human before they are incorporated into a final manuscript.
• Authors must correct inaccuracies and fabricated facts (‘hallucinations’) and ensure all claims are cross-checked against peer-reviewed sources. 

Edinburgh University Press (or journal/series editors) may reject manuscripts if an author has not followed our AI policy. 

Definitions

• Research Integrity: A commitment to responsible and ethical uses of AI in such a way as to ensure honesty, accuracy and accountability in all research stages.
• Ethical Use of AI: An approach to AI which avoids plagiarism, copyright infringement, data fabrication or misrepresentation of research, while ensuring that AI-generated content respects copyright, privacy and human oversight while minimising bias and harm. (e.g., using AI for grammar checks with disclosure is ethical, but generating conclusions without appropriate citations is not).
• Disclosure: The explicit acknowledgement of the AI tools used, the purposes for which they have been used, the outputs that have been produced and the verification steps that have been taken.
• Generative AI: This is a type of AI that creates new content based on what it has learned from existing training data. 
• LLM (large language model): A type of generative AI that uses a set of training data to generate new written content (e.g., ChatGPT). LLMs will use your inputs and interactions with it as training data to learn from.
• Generative image model: A generative AI that uses a set of training data to generate new visual content (e.g., DALL-E).
• AI chatbot: The interface into which users input a prompt and receive an AI-generated output. These chatbots will be powered by a form of AI technology, such as a large language model.
• Walled garden model: An AI model with a closed dataset, meaning that anything uploaded to the model is only accessible to its users. Some academic institutions have created walled garden models for their students and staff to use as a more secure alternative to commercial models.

Disclosure and transparency

As AI tools become increasingly embedded in the research process, it is crucial to be transparent about their use. Open disclosure promotes research integrity, fosters trust and ensures compliance with ethical, institutional and publishing standards.

Any use of AI must not breach EUP’s publication ethics. Scholarly works must be the author’s own and must not present third-party ideas, data, words or other material without proper citation and transparent referencing.

Authors are responsible for the accuracy of all information in their work. This includes fully disclosing the use of AI tools and verifying any information or citations they have generated.

We will therefore require authors to submit an AI Disclosure Form with their submissions. Edinburgh University Press will then review whether the use and rationale provided aligns with our policy.

When disclosing AI use, authors must:

• Include it in the acknowledgements section.
• Specify the name and model of the AI, as well as the date it was accessed.
• Explain how and why it was used in your research process.
• Maintain documentation of all AI-generated/AI-assisted technology used, and how you personally reviewed and verified AI-generated content.
• If use of AI is extensive, Edinburgh University Press may ask authors to provide an AI appendix with further details of how generative AI was used.

If reviewers, editors or authors suspect use of AI in our work that has not been declared, they will inform Edinburgh University Press. These instances will be fully investigated in line with COPE guidance, and appropriate action will be taken. This may include rejecting the book, chapter or article, issuing a correction, retraction or withdrawal from sale, and/or informing authors’ institutions.

Sample disclosure statement

We may ask authors who have used AI tools to insert a statement at the beginning of a book manuscript or at the end of a journal manuscript above the references or bibliography. The statement should be entitled ‘Declaration of AI technologies in the writing process’. 

In that statement, authors must specify the AI (including version/model and date accessed) and the reason for which it was used. We suggest that authors follow this format when preparing their statement:

During the preparation of this work, the author(s) used [NAME AI & VERSION OF THE MODEL] in order to [REASON]. After using this service, the author(s) verified the output and take(s) full responsibility for the content of the publication.

Use of Assistive AI

We recognise that AI-assisted writing has become more common as the technology becomes more accessible. As such, we allow the use of assistive AI tools that make suggestions to improve the language, grammar or structure of your work (e.g., Grammarly). 

Assistive use of AI must still be disclosed upon submission and authors should not submit their whole unpublished manuscripts to any AI chatbot for assistive purposes, as this could compromise their intellectual property rights.

• Authors may use AI for the purpose of editing on a limited and assistive basis. 
• AI may assist in improving grammar, readability or non-substantive phrasing (e.g., rephrasing awkward sentences).
• Authors may use commercial AI tools to edit excerpts of their manuscript, however uploading the full unpublished work to these tools may compromise copyright and is therefore prohibited. 
• We encourage authors to check the AI’s settings to ensure that the prompts and work submitted will not be used as training data.

Authors will also still be responsible for verifying any AI-generated output and ensuring that AI-assisted content does not compromise research integrity. 

Additionally, please note that when information is entered into generative AI, the organisation which owns the generative AI tool will likely have access to this data. Please be aware of this possibility and how it may impact the privacy of any research participants, as well as how it may impact author privacy and intellectual property. 

For this reason, authors, editors and/or reviewers must not enter materials from submitted manuscripts into generative AI as it would constitute a violation of the confidentiality of the peer review process.

If authors do submit parts of their manuscript to an LLM, it must be a local, closed or ‘walled garden’ model (e.g., ELM) to protect author copyright. We recommend asking your institution if they have a walled garden model. 

If authors use AI to produce or amend their content rather than their language, then that will no longer be considered assistive use of AI.

Idea generation

• Authors should ensure they are not relying solely on AI for idea generation without human validation of scholarly merit. At present, generative AI lacks the ability to adhere to the standard scholarly community practice of accurately attributing ideas to the original author, therefore ideas presented by AI must be thoroughly cross-checked and verified. However, authors must disclose this and ensure they adhere to research integrity and ethics.
• A key cornerstone of scholarly publication is that the work is original. Any use of AI for idea generation must not compromise this.

Peer review

• Reviewers are prohibited from using AI tools to generate a peer review. Again, they must not upload any unpublished manuscripts to AI tools, as this could breach copyright and confidentiality.
• Assistive AI tools may be used to edit the language or tone used in a review; however, any identifiable or confidential information must not be uploaded, and reviewers should only upload excerpts of their review. 
• Any AI output must be verified by the reviewer, and they will be liable for any errors found in the review.

Citations

Authors must disclose whether they have used AI to create a list of sources for their manuscript, confirm that they have checked all sources and cite the original source, not the AI model. Regardless of how AI was used in the process of developing an article, it cannot be cited as a source. 

Any sources that are provided by AI need to be verified by the author. This includes both ensuring that the sources exist and that their content has been accurately reflected in the AI output.

Translation

While AI can be a very useful tool for translation, Edinburgh University Press does not allow authors to use AI tools to translate whole manuscripts, including full books chapters or articles. Generative AI tools should not be used for translation purposes and Assistive AI tools should only be used on a limited basis. For example, it would be acceptable to use an assistive tool (such as Grammarly or DeepL) to translate a particularly difficult paragraph. However, authors are not allowed to use generative tools (such as ChatGPT) for this purpose or to put full sections of the manuscript into assistive tools.

We encourage authors to check the AI’s settings to ensure that the prompts and work submitted will not be used as training data. We would ask authors to use human translators wherever possible for accuracy and safety.

Images

AI should not be used to generate images for either internal or cover use. The only exception is if AI-generated images are specifically discussed in the manuscript and you are providing a visual example.

Similarly, authors must not download or purchase AI-generated images for use. Our recommended stock photo agencies – Getty Images, Shutterstock, iStockphoto, Superstock – do not accept AI-generated submissions to their databases but other agencies might. 

Edinburgh University Press and our design partners will not use generative AI to create book or journals cover designs. Assistive AI may be used to make minimal alterations such as adding colour washes or extending backgrounds. 

AI can be used to help with the presentation of data or processes (in charts or flowcharts for example) so long as that data cannot be used in training the AI model.

AI can be used for the generation of alt text, so long as the image or book cover cannot be used in training the AI model.

Promotion and book reviews

Edinburgh University Press permits some use of AI to help you promote your book.

AI for blog posts

Are you writing a piece for the Edinburgh University Press Blog or any other online or offline news or opinion outlet?

While we do not permit use of generative AI for writing blog posts, assistive AI can be a great tool for generating ideas and outlines, headlines and keywords, alt text for images and blog post summaries.

When using AI to help with your blog post/news article, it’s important to remember:

• AI often lacks nuance and can get things wrong, including misquoting sources and generating outdated information.
• AI-generated writing can be too broad or read on the hyperbolic side, so you will often find that it doesn’t match your natural voice.
• Do not enter any personal, copyrighted or confidential information into AI tools (e.g., articles, chapters, existing unposted blog posts, content from other creators, copyrighted images, data).
• Always use human verification, including to proof your blog submission. AI tools such as Grammarly are helpful in assisting with initial editing, but human editors should always do the final review to ensure factual accuracy.
• Be wary of free AI tools and always read the fine print. Anything you input into an AI tool is usually used to help train it.

AI for social media posts

AI can be a great tool to help you with social media posts. We recommend using it for wordcount suggestions (e.g., if your post is too long, and you need help cutting it down), hashtag ideas and alt text for your images. It is not recommended to use AI to generate social media posts unless they are checked by a human for accuracy and grammar.

AI for book reviews

We do not permit the use of AI to generate book reviews for our titles or for reviews of other books within EUP journals. Moreover, the uploading of digital book files into AI is strictly prohibited. Reviews should be honest, transparent and human. Always check with your review editor on their individual policies on AI (for example, for abstract and summary generation, keywords, idea generation).

Crib sheet: Use of AI