AI Use in Manuscript Preparation

Using AI safely when preparing your manuscript

Artificial intelligence (AI) tools are increasingly used to support manuscript preparation, from improving clarity and structure to assisting with drafting and summarisation. These uses can enhance communication and accessibility, but they vary in the level of influence they have on the scholarly content.

91죤펜 applies a risk-assessment approach, focusing on how AI is used, rather than whether it is used.

A risk-assessment framework for AI use in writing

Risk level	What this means	Typical use in manuscript preparation	Key considerations	Declaration
Green (low risk)	AI supports clarity and presentation without altering scholarly content	Grammar and language editing; improving readability and flow; formatting; translation; summarising author-written text	Outputs must be checked for accuracy and meaning	Clearly describe use and confirm author accountability
Amber (requires caution and care from authors)	AI contributes to structuring or shaping written content	Generating outlines; drafting sections based on author input; suggesting alternative phrasing or framing; reference management support; creating visual content with verifiable source data or material	Risk of generic or inaccurate content, loss of disciplinary voice, fabricated or incorrect citations, all AI generated or assisted content must be clearly labelled	Clearly describe use and confirm author accountability
Red *(not permitted)*	AI replaces authorship or generates scholarly content without oversight	Writing results or discussion independently; generating scientific claims or interpretations; producing data or figures as if empirical; listing AI as an author; creating content from text prompts or without verifiable source data or material	Undermines originality, accuracy, and accountability; risk of infringing copyright, licensing, or other intellectual property rights	Not permitted

What this means in practice

Many uses of AI in writing are low risk and widely accepted, for example, improving grammar, clarity, or structure. These uses are surface-level, reversible, and remain fully under author control.

More extensive uses, such as drafting or restructuring content, require greater care. Authors should ensure that:

all content is accurate and verified
disciplinary conventions are correctly applied
references are real and have been checked
the intellectual contribution remains their own

AI can assist with expression, but the argument, interpretation, and conclusions must be author-led.

AI-generated and AI-assisted visual content

AI may be used to create, modify, enhance or analyse visual content where its use is transparent, lawful, and does not compromise the accuracy or integrity of the scholarly and scientific record.

For the purposes of this policy, visual content includes figures, graphs, charts, maps, diagrams, illustrations, photographs, images and video content.

Visual content derived from verifiable inputs and source material

The use of AI to generate, analyse, visualise or enhance visual content may be permitted where authors meet the following requirements:

AI use is disclosed appropriately, including model and purpose of use;
the underlying data, source material, methods or computational outputs can be inspected, verified and are clearly attributed;
any AI-assisted processing, analysis or code generation has been subject to appropriate human review and validation;
the resulting visual output accurately represents the underlying evidence;
authors remain responsible for ensuring that the use of AI-generated or AI-assisted visual content does not infringe copyright, licensing or other intellectual property rights.

Examples may include the creation of charts from datasets, visualisation of computational outputs, diagrams derived from author-developed content, scientific images created from attributable data sources, or figures generated using code that has been reviewed and verified by the author.

Visual content without verifiable inputs and source material

AI generated visual content that is not derived from independently verifiable data, source material, methods, computational outputs or author-developed content is considered opaque and is not permitted by 91죤펜 policy.

Authors, reviewers and editors should be able to assess how the content was created and whether its accuracy, authenticity and legal status can be reasonably established.

Non-generative AI tools

Not all AI tools are generative.

The use of machine learning tools to analyse, improve, combine, restore, segment or enhance existing visual content may be permitted where the underlying material remains attributable and the changes do not misrepresent the original content. Such use should be disclosed for editorial assessment.

Transparency and declaration

Some authors may hesitate to disclose AI use due to concerns about how it might be perceived. However, the policy is designed to support openness and consistency.

Declaration does not negatively influence editorial decisions
It enables transparent and fair evaluation
It reduces the need for clarification during editorial assessment

Where AI use is not declared, editors may need to investigate or request clarification, which can delay the review process and shift attention away from the manuscript�셲 contribution.

Declaring AI use helps ensure that your work is assessed on its scholarly merit.

Transparency and disclosure for visual content

Visual content that has been created, substantially modified or enhanced using AI should be disclosed clearly and transparently.

Where AI-generated or AI-assisted visual content is published, information about the AI system used, the purpose of its use, and the extent of its contribution should be provided in an AI Declaration in your manuscript.

Specific disclosure should be provided within the caption or legend of the visual content.

Key principle

AI is safe when it helps you communicate your research. It becomes unsafe without adequate human oversight and accountability.

You remain fully accountable for everything submitted in your name.

91죤펜