“Visual Sexual Health Communication Assisted by AI: The Case of Synthetic Menstruation Images” (2026), by Nicola M. Döring et al. Computer Science research paper. Original abstract below, unchanged. A free PDF is available.

Visual Sexual Health Communication Assisted by AI: The Case of Synthetic Menstruation Images

Nicola M. Döring, Anastasiia Shevtsova · November 3, 2026 · Computer Science

Abstract

Public communication about sexual and reproductive health and rights (SRHR) relies on both text and images. A recent development is the proliferation of artificial intelligence (AI) image generators, which enable rapid, cost-effective, and highly customizable images with appealing aesthetics based on textual requests (prompts). To examine the opportunities and challenges of AI-generated images in SRHR communication, we used the topic of menstruation and developed six research questions (RQs) anchored in the current state of research on AI-supported visual health communication. These questions addressed key areas of concern, such as unsuccessful image creation (RQ1), objective image flaws (RQ2), lack of diversity of human representation (RQ3), stereotypical portrayals of women (RQ4), use of visual symbols (RQ5), and image valence (RQ6). Our study applied 15 AI image generators, 2 user accounts per generator, 5 menstruation-related prompts, 2 prompt languages, and 2 images per prompt, resulting in a sample of N = 600 AI outputs generated in Autumn 2025. All outputs were manually coded using a reliability-tested codebook developed through both deductive and inductive approaches. In 20.0% of cases, menstruation prompts did not produce the requested image but only text output (RQ1). When images were generated, they often (78.1%) exhibited visual and/or textual flaws (RQ2). The images showed a lack of diversity in some human attributes (e.g., body type) but not in others (e.g., skin tone; RQ3). Stereotypical depictions of women were observed but not predominant (RQ4). Common menstruation-related visual symbols (e.g., blood, pain) were identified (RQ5), but most images did not convey a negative portrayal and instead a neutral or positive portrayal of menstruation (RQ6). While results show that initial outputs from AI image generators are often not directly suitable for visual sexual health communication, they might work with targeted refinement. Further research is required to examine the production, characteristics, consumption, and effects of AI-generated images in SRHR communication.

Free PDF DOI: 10.1007/s10508-026-03518-x