03 · What You Need to Know
Understanding Disclosure Requirements for AI-Generated Graphical Abstracts
Why Does AI Use in Graphical Abstracts Raise Disclosure Questions?
A graphical abstract communicates the main contribution of a research paper through visual elements. Readers may encounter it in a journal's online contents, article page, or promotional materials before reading the full manuscript.
Because the image represents scholarly work, its production can raise questions about accuracy, attribution, originality, and accountability. These concerns become particularly relevant when an AI system generates visual content that the researchers did not draw or directly construct themselves.
Disclosure can help editors and readers understand how the visual was produced and what role the authors played in verifying it. However, transparency should not be confused with evidence that the content is scientifically valid.
The responsibility for the final graphical abstract remains with the authors, even when an AI tool generated substantial portions of the image.
What Counts as AI Assistance in a Graphical Abstract?
AI involvement can range from suggesting a visual arrangement to generating the complete illustration. These activities may be treated differently under publisher policies.
| Type of Assistance |
Example |
Disclosure Consideration |
| Conceptual planning |
AI proposes a storyboard based on the manuscript |
Check whether the publisher requires reporting substantive AI assistance during preparation |
| Text refinement |
AI shortens labels or improves wording |
Apply the publisher's AI-assisted writing and editing rules |
| Layout assistance |
AI suggests the placement of researcher-created elements |
Determine whether the tool's contribution falls within reportable visual assistance |
| Generated illustrations |
AI creates icons, objects, or scientific scenes |
Verify permission and any required disclosure of generated artwork |
| Complete image generation |
An image generator produces the entire graphical abstract |
Check explicit prohibitions before considering disclosure |
| AI-assisted modification |
AI adds, replaces, or substantially transforms visual elements |
Assess whether the modifications are permitted and must be documented |
The relevant distinction is not simply whether AI was used. It is what the system contributed, how the resulting content was produced, and whether the publisher permits that contribution.
What Do Elsevier's Policies Say About AI-Generated Graphical Abstracts?
Elsevier's updated generative AI policies for journals distinguish among explanatory figures, research images, graphical abstracts, and cover artwork.
Under its graphical abstract provisions, general-purpose generative AI image tools must not be used to create graphical abstracts. Elsevier instead directs authors toward dedicated scientific illustration tools or other professional illustration software, subject to the applicable requirements.
Its policy also identifies disclosure expectations for permitted scientific illustration tools, including mentioning the tool used in the image caption.
This creates an important distinction: an author cannot make a prohibited AI-generated graphical abstract acceptable simply by declaring that an image generator was used.
Watch Out
If a publisher prohibits a particular method of generating graphical abstracts, adding an AI disclosure statement does not override that prohibition. Permission and disclosure are separate requirements, and both must be addressed before submission.
Does Springer Nature Apply the Same Requirements?
Not exactly. Springer Nature's AI manuscript preparation policy allows certain AI-assisted visual content when its creation is transparent, lawful, and does not compromise the scholarly record.
For visuals derived from verifiable data, source material, methods, computational outputs, or author-developed content, the policy specifies conditions concerning disclosure, human validation, attribution, and intellectual property.
It also distinguishes such content from opaque AI-generated visuals that lack independently verifiable inputs or source material.
These provisions should not be interpreted as unrestricted permission to generate graphical abstracts. Researchers must establish whether their particular visual and production method satisfy the policy and any additional journal requirements.
The difference between these publisher approaches illustrates why a single universal disclosure statement cannot substitute for checking the intended publication venue.
Where Should AI Assistance Be Disclosed?
The appropriate location depends on the journal's instructions and the type of assistance involved.
Possible locations include an AI-use declaration, the graphical abstract's caption, a relevant acknowledgment, or the submission system's disclosure fields. Where AI use is part of the research methodology rather than manuscript preparation, additional reporting in the Methods section may be necessary.
Researchers should avoid assuming that a declaration in one location satisfies every requirement. A publisher may request a specific disclosure format or distinguish AI-assisted writing from AI-generated visual content.
When a graphical abstract is submitted as a separate image file, authors should also determine how the required information will accompany that file. Adding a disclosure statement directly inside the image may be inappropriate if the journal prohibits additional text or requires a particular layout.
What Information Should an AI Disclosure Include?
When disclosure is required, it should accurately describe the contribution rather than merely announcing that artificial intelligence was used.
Depending on the publisher's instructions, relevant information may include the tool's name, its developer, the version where applicable, the purpose of its use, the nature of the generated or modified content, and the verification performed by the authors.
For example, there is a meaningful difference between using a language model to propose a layout and using an image generator to create a complete scientific illustration.
A disclosure should make that distinction understandable.
Insufficiently specific
Artificial intelligence was used to prepare the graphical abstract.
More informative
Identifies the tool, explains whether it assisted with planning, wording, layout, or image generation, and describes the authors' review of the resulting content.
Greater detail is not always necessary. The appropriate level of reporting should follow the publisher's requirements and the materiality of the AI contribution.
Does Using AI Only to Suggest a Layout Require Disclosure?
Not necessarily under every policy. Some publishers distinguish routine editing or preparatory assistance from substantive AI-generated content.
If a researcher uses AI to brainstorm possible arrangements but manually creates the final graphical abstract, the disclosure requirements may differ from those applying to AI-generated artwork.
However, the absence of generated pixels does not automatically mean that the contribution is irrelevant. AI might influence which results are emphasized, how relationships are represented, or which conclusions appear most prominent.
Researchers should therefore assess both the technical nature and intellectual significance of the assistance.
What If AI Generates Only One Small Illustration?
The size of an AI-generated element does not necessarily determine whether it is permitted or reportable.
A single AI-generated scientific illustration can still raise concerns about originality, licensing, or scientific accuracy. If the publisher prohibits general-purpose AI-generated graphical abstract artwork, using such a tool for only one component may still conflict with its policy.
Researchers should not attempt to avoid disclosure by combining generated elements with manually created ones. The relevant issue is the actual production process, not the proportion of the final image occupied by AI-generated content.
Is Disclosing AI Use the Same as Crediting the AI as an Author?
No. Disclosure describes the contribution of a tool or system. Authorship concerns responsibility and accountability for scholarly work.
AI systems cannot assume the responsibilities associated with human authorship, including verifying the research, approving the final publication, and responding to concerns about accuracy or integrity.
Researchers should identify the AI contribution where required without presenting the system as an accountable scholarly author.
What Records Should Researchers Preserve?
When AI materially contributes to a graphical abstract, retaining a record of the production process can help demonstrate how the final image was developed.
Depending on the workflow, useful records may include the original researcher-approved storyboard, relevant prompts, tool and version information, generated drafts, manually edited files, and the final submitted image.
These records can also help distinguish legitimate design changes from alterations that introduce unsupported scientific claims.
The documentation need not be unnecessarily elaborate. Its purpose is to make the production process understandable and support verification if questions arise during editorial assessment.
04 · A Practical Example
Deciding Whether and How to Disclose AI Assistance
Hypothetical Example
A Graphical Abstract for an AI-Assisted Learning Study
A researcher has completed a study examining students' experiences with an AI-supported learning platform. The researcher wants a graphical abstract showing the instructional setting, the study's approach, and the principal findings.
The researcher first uses a language model to propose a visual storyboard. The final illustration is then constructed manually using editable shapes and appropriately licensed graphical elements.
Before submission, the researcher discovers that the target journal distinguishes AI-assisted preparation from AI-generated artwork.
Step 1: Identify the actual AI contribution
The researcher records that AI assisted with the preliminary storyboard but did not generate the final illustrations or independently determine the research findings.
Step 2: Examine the publisher's policy
The researcher checks whether AI-assisted planning is reportable and confirms that the selected illustration method complies with the journal's graphical abstract requirements.
Step 3: Verify the scientific content
The researcher compares the final graphical abstract with the manuscript and ensures that its visual claims reflect the study's findings and limitations.
Step 4: Prepare an appropriate disclosure
If the policy requires disclosure of the planning assistance, the researcher identifies the tool and describes its role in proposing the initial layout. The statement does not imply that the AI generated the final artwork.
Step 5: Submit the required documentation
The researcher places the disclosure in the location specified by the journal and retains the relevant production records.
A possible disclosure, where such wording is consistent with the journal's requirements, would be:
Illustrative Disclosure
"During the preparation of the graphical abstract, the authors used [tool name and version] to explore preliminary layout options based on an author-prepared summary of the study. The final graphical abstract was manually constructed and reviewed by the authors, who take responsibility for its scientific accuracy and presentation."
This statement should not be copied without adjustment. It is appropriate only if it accurately describes the actual workflow and the publisher permits that workflow.
If the researcher had instead used a general-purpose image generator to produce the complete graphical abstract for a journal that prohibits such artwork, the correct response would be to replace the image with one produced through an acceptable method, not merely to add a disclosure.
