AI-Generated Figures and Text on Scientific Posters: What Congresses and Journals Allow, and How to Disclose It

As generative AI research tools become routine, many authors use ChatGPT and other large language models to polish wording, and image generation tools to create icons or schematics. Using AI in scientific posters raises two questions: what is allowed, and how should it be disclosed?

A poster usually becomes a published abstract and later a journal article, so getting AI generated figures disclosure right now avoids problems later.

AI use on a poster falls into four categories, and each is treated differently (Table 1). Basic spelling and grammar tools are often exempt from disclosure, whereas generated text, generated images and AI used in the research itself usually need a statement. The exact threshold depends on the congress or journal.

The ICMJE AI guidance, in the recommendations updated in January 2026, sets the core position for medical journals [1]. Chatbots such as ChatGPT cannot be authors because they cannot be responsible for the accuracy, integrity and originality of the work. Authors should describe AI use in both the cover letter and the submitted work: writing assistance goes in the acknowledgments, and AI used for data collection, analysis or figure generation goes in the methods, with enough detail to replicate it, including tool, version and prompts where applicable. Authors must review AI output and be able to assert that AI-produced text and images contain no plagiarism [1].

The Committee on Publication Ethics (COPE) takes the same authorship policy position: AI tools cannot meet authorship requirements because they cannot take responsibility for the work. Authors who use AI to write text, produce images or graphical elements, or collect or analyse data must disclose how and which tool was used in the methods or a similar section, and remain fully responsible for the content [2].

Beyond these shared principles, what AI disclosure journals require varies. A bibliometric analysis found that 87% of the top 100 journals had guidance on generative AI and 98% of those prohibited AI authorship [3]. Image rules diverged: Science prohibited AI-generated text, figures, images and graphics; Emerald and BioMed Central prohibited AI-generated images; and Blood allowed graphical but not textual AI output. Of 15 journals with both their own and their publisher’s guidance, 12 conflicted [3]. Check the target journal’s current instructions; see also our publisher AI disclosure guide.

Congress rules on AI vary and mostly address the abstract, not the poster. The European Society for Medical Oncology (ESMO) is an example: its 2025 Immuno-Oncology abstract regulations allow generative AI in writing an initial abstract draft, with human oversight; authors must review and edit the final abstract and accept full responsibility, and AI may not be listed or cited as an author [4]. AI used to generate or analyse research data must be described in the methods of the abstract or presentation. Basic grammar, spelling and reference-checking tools need no declaration, and the presenting author must have full control over the abstract and the subsequent slides, poster or ePoster [4].

Image generators can produce realistic but fictitious images, so a generated image must never be presented as study data. Microscopy images, blots, scans, clinical photographs and data plots must come from the research itself.

Illustrative graphics, such as a study-design schematic or mechanism diagram, are treated differently: some journals prohibit them, others allow them with disclosure [3]. Good AI image disclosure means labelling every AI-generated graphic in its caption, naming the tool and version, stating that it does not represent study data, and keeping the original files. Also check that the tool’s terms permit reuse in a publication, because licensing conditions vary.

Generated text can read fluently while lacking substance. In one study, blinded human reviewers identified 68% of ChatGPT-generated abstracts but misclassified 14% of real ones, describing suspected generated abstracts as vaguer and more formulaic [5]. Check AI-assisted poster text for accuracy, specific numbers and correct citations, because errors remain the authors’ responsibility.

• Abstract submission: answer any AI questions on the submission form.
• Acknowledgments or footnote: add a short transparency statement for writing assistance.
• Figure captions: label every AI-generated or AI-edited graphic.
• Methods box: describe AI tools used in data collection or analysis, with version and purpose.

A statement can be brief: “The authors used [tool, version] to improve the readability of the poster text. The authors reviewed and edited all content and take full responsibility for it.” For a caption: “Schematic created with [tool, version] and edited by the authors; it does not represent study data.” Record the tool, version, date and prompt for any substantive output.

• Check the congress rules and your target journal’s policy on AI use and AI-generated images.
• Confirm no AI tool is listed as an author and all authors approve the final content.
• Verify every number, claim and citation in AI-assisted text against your data and sources.
• Confirm all data images and plots come from the study, not an image generator.
• Label AI-generated illustrations and add a transparency statement where required.
• Save a record of tools, versions and prompts for the journal submission.

For layout and design, see our step-by-step research poster guide, and for the abstract behind the poster, our guide to why conference abstracts get rejected.

AI can support poster preparation, but it does not change who is accountable. Language editing usually needs little or no disclosure, generated text and illustrative graphics need a clear transparency statement, and generated images must never stand in for real data. Check both before you design the poster.