03 · What You Need to Know
When and How Researchers Should Disclose AI-Generated Presentation Graphics
Why Does Disclosure Matter in Research Presentations?
Research presentations communicate evidence, interpretations, and scientific arguments. Graphics influence how audiences understand these claims.
When an image is generated by AI, its appearance may not correspond to an actual observation, participant, instrument, or research setting.
Disclosure can help audiences distinguish illustrative material from empirical evidence and understand how particular scientific representations were produced.
However, disclosure is not a substitute for accuracy.
A scientifically incorrect diagram remains problematic even if it is clearly labeled as AI-generated. Similarly, a generated image should not be presented as an authentic research photograph merely because its source is disclosed elsewhere.
Transparency and scientific verification serve complementary purposes.
Are Researchers Always Required to Disclose AI-Generated Images?
No universal requirement applies to every academic presentation in every institution or discipline.
Disclosure obligations depend on the applicable conference rules, institutional policies, publisher requirements, professional standards, and circumstances of use.
Some policies specifically address generative AI images. Others focus on manuscript preparation, research data, or substantive AI contributions.
Researchers should not assume that a policy governing journal manuscripts automatically applies unchanged to conference slides.
Conversely, the absence of an explicit conference rule does not eliminate broader obligations concerning research integrity, confidentiality, copyright, or truthful representation.
When the applicable policy is unclear, a concise and accurate disclosure may help avoid ambiguity, particularly for substantive AI-generated scientific content.
Does the Purpose of the Graphic Affect Disclosure?
Yes. The role of the image is central to determining what information audiences need.
| Graphic type |
Example |
Disclosure consideration |
| Decorative illustration |
Generated image of a hypothetical classroom |
Follow applicable rules; consider whether its illustrative status is obvious. |
| Conceptual illustration |
Generated visual metaphor for AI-assisted learning |
Identify AI generation when required or when provenance is relevant. |
| Scientific diagram |
AI-generated theoretical framework |
Document substantive AI involvement and verify all relationships. |
| Methodological diagram |
AI-generated research workflow |
Verify procedures and consider disclosure of consequential generation. |
| Data-bearing figure |
AI-generated or modified statistical graph |
Preserve traceability to the data and comply with figure-processing policies. |
| Synthetic research image |
Generated example used in an experiment |
Describe its generation as part of the research methodology when relevant. |
The appropriate disclosure depends not only on whether AI was used but also on what the audience might reasonably infer from the image.
Should Decorative AI Images Be Disclosed?
Decorative images may have a different disclosure significance from graphics that communicate research findings.
For example, a generated illustration of a student studying with a laptop may simply establish the presentation's topic.
If the image is clearly illustrative and does not communicate empirical information, the scientific consequences of its generation may be limited.
Nevertheless, conference or institutional policies may require disclosure regardless of the image's purpose.
Researchers should check those requirements and consider whether a brief acknowledgment would improve transparency.
Disclosure need not imply that the generated illustration is itself research evidence.
Should AI-Generated Scientific Diagrams Be Disclosed?
Scientific diagrams can communicate substantive information through components, labels, arrows, and relationships.
When AI generates or substantially modifies these elements, disclosure may be relevant because the tool has contributed to the representation of scientific content.
However, the researcher must still verify that the diagram accurately reflects the intended framework, protocol, or scientific model.
For example, an AI-generated conceptual framework may introduce a causal arrow that the original study did not support.
Disclosure would inform audiences about the tool's involvement, but it would not correct the misleading relationship.
Researchers should therefore address the risks of inaccuracies in AI-generated scientific diagrams separately from the question of disclosure.
What If AI Only Improves the Appearance of an Existing Figure?
The significance of AI involvement depends on the modification.
Changing font sizes, adjusting spacing, or applying a verified color palette may be primarily presentational.
However, generative editing can also reconstruct labels, alter data points, or change uncertainty intervals.
Researchers should establish whether the tool modified only presentation properties or affected the scientific representation.
Applicable publisher and conference policies may distinguish routine image adjustments from generative modifications.
Regardless of disclosure requirements, the final figure should be checked against its source evidence.
Can AI-Generated Images Be Mistaken for Research Photographs?
Yes. Photorealistic generated images may resemble authentic photographs of participants, laboratories, classrooms, or experimental settings.
If presented without context, audiences may assume that the image documents an actual research observation.
For example, a generated image showing students using a particular educational technology platform should not be presented as a photograph from the study unless it genuinely documents the research setting.
Researchers should label hypothetical or illustrative images when necessary to prevent confusion.
This distinction is particularly important when the visual appearance of the research setting influences how audiences interpret the study's authenticity or implementation.
What If AI-Generated Images Are Part of the Research Methodology?
When generated images are used as experimental stimuli, research materials, or objects of analysis, their provenance may be methodologically essential.
For example, a study examining students' ability to distinguish authentic from AI-generated educational images would need to describe how the generated materials were produced.
Relevant details may include the tool, available model identifier, prompts, generation settings, selection criteria, and any postprocessing that affects interpretation or reproducibility.
The appropriate level of documentation depends on the research question and methodological requirements.
In this context, AI-image provenance is not merely an optional presentation acknowledgment. It may be necessary for evaluating and reproducing the study.
Does Disclosure Require Naming the AI Tool?
Some policies may require identifying the tool, provider, or model. Others may request only a general description of the assistance.
Researchers should follow the applicable requirements and avoid implying a level of technical specificity that they cannot verify.
For example, a disclosure may identify that a conceptual illustration was generated using a particular image-generation service and subsequently reviewed or edited by the authors.
If the precise model version is unavailable, researchers should not invent one.
Where model identity is relevant to the research methodology, available technical information should be preserved in the research record.
Where Should Disclosure Appear?
The appropriate location depends on the presentation format and applicable policy.
For a generated illustration, a concise image credit or note may be suitable when disclosure is needed.
For a presentation containing several generated graphics, a general acknowledgment slide may be appropriate if it identifies the relevant material clearly.
For AI-generated experimental stimuli or methodological figures, the research methods or accompanying documentation may need more detailed information.
Researchers should avoid disclosures so vague that audiences cannot determine which graphics were generated or modified.
At the same time, disclosure should not overwhelm the presentation or distract unnecessarily from the scientific content.
What Might an Appropriate Disclosure Look Like?
The wording should describe the actual use rather than rely on a generic statement.
For a decorative image, a concise acknowledgment might state that the illustration was generated using an AI image tool and does not depict research participants.
For a conceptual diagram, the disclosure might explain that AI assisted with graphical layout while the scientific relationships were specified and verified by the authors.
For research stimuli, the methods may need to describe the generation process in sufficient detail for methodological evaluation.
Researchers should not claim that an image was fully human-created if AI generated substantial visual content.
Similarly, they should not describe a data-driven figure as AI-generated merely because AI helped write the plotting code, unless that description accurately reflects the relevant policy and workflow.
Is AI-Assisted Plotting Code the Same as AI-Generated Imagery?
Not necessarily.
Suppose AI helps write Python code that plots verified regression coefficients from an analytical dataset.
The resulting figure is generated computationally from specified numerical values, rather than synthesized through a generative image model.
The code may still contain errors, and AI assistance may require disclosure under applicable policies.
However, the distinction matters because the risks and provenance differ.
A data-driven graph should remain traceable to its analytical inputs, while a generated illustration may contain synthesized visual details that do not correspond to empirical observations.
Researchers should describe the actual process rather than treat every AI-assisted graphic as technically equivalent.
Can Disclosure Resolve Copyright Concerns?
No. Identifying AI involvement does not automatically establish that the resulting graphic can be lawfully used or distributed.
Copyright and licensing questions may depend on jurisdiction, contractual terms, source materials, and the nature of the output.
For example, an AI-generated graphic that closely resembles a protected illustration may raise issues that are not resolved by adding an AI disclosure.
Similarly, uploading a copyrighted figure to an external AI service may involve permissions or contractual restrictions.
Researchers should evaluate rights and licensing separately from transparency requirements.
Can AI-Generated Graphics Create Misleading Scientific Impressions?
Yes. A generated image may suggest that a particular procedure occurred, that participants had certain characteristics, or that a scientific mechanism has been established.
These implications may arise even when the image contains no explicit numerical claims.
For example, a generated laboratory illustration might depict equipment that was not used in the study.
Researchers should examine whether the graphic's visual content could reasonably be interpreted as evidence about the research.
Where confusion is possible, labeling the image as illustrative may be necessary, but misleading scientific details should also be corrected.
Should Researchers Preserve AI-Generated Graphics and Prompts?
Preservation requirements depend on the role of the graphics and applicable research policies.
For decorative illustrations, the final image and relevant licensing or attribution information may be sufficient in some contexts.
For graphics used as research stimuli or methodological materials, prompts, model information, selection procedures, and original outputs may be necessary for reproducibility.
Researchers should distinguish between documentation needed for scientific evaluation and records retained for ordinary presentation development.
Where AI substantially modifies scientific diagrams, preserving the original and final versions can help establish what changed.
What If the Conference Has No AI Graphics Policy?
The absence of an explicit rule does not automatically prohibit or authorize every use.
Researchers should consider institutional requirements, research integrity standards, copyright conditions, and whether the audience could misunderstand the graphic's provenance.
A concise disclosure may be prudent when AI-generated content plays a substantive role in representing scientific information.
For purely decorative material, the appropriate approach may depend more heavily on context and local expectations.
Researchers should avoid claiming that disclosure is universally mandatory when no applicable requirement has been established.
Does Disclosure Guarantee Research Integrity?
No. Disclosure communicates information about how content was produced, but it does not establish that the content is accurate.
A generated scientific diagram may contain incorrect relationships even when its AI origin is clearly acknowledged.
A data-driven figure may contain incorrect statistical values despite complete documentation of the plotting workflow.
Researchers should therefore combine transparency with substantive verification.
The broader process of verifying AI-assisted research presentations should address all scientific content, not only whether AI involvement has been disclosed.
