Generative artificial intelligence (AI) is increasingly being applied in intimate and sexuality-related contexts, such as the creation of couple images or the use of AI-supported dialogue systems in couple and sexual counseling. At the same time, such applications are often met with skepticism. One well-documented phenomenon in this context is anti-AI bias, referring to the tendency to evaluate AI-generated content more negatively than equivalent human-created content, regardless of its actual quality.
A recent study by the Media Psychology and Media Design Group at TU Ilmenau examined this effect in two consecutive online experiments with a nationally representative sample from Germany (N = 2,658). Participants were presented with identical romantic couple images and excerpts from sexuality-related couple counseling, which were labeled either as AI-generated or as human-created. The results show that both types of content were evaluated significantly less positively when described as AI-based.
In addition, the study analyzed whether individual attitudes toward AI and levels of AI literacy influenced these evaluations. While a moderating effect was observed for the evaluation of couple images, no such effect emerged for the counseling dialogues. Overall, the findings extend research on anti-AI bias to the previously underexplored domain of intimacy and sexuality and provide important insights for the design and acceptance of generative AI systems in sensitive application areas.
Döring, N., & Mohseni, M. R. (2025). Anti-AI bias toward couple images and couple counseling: Findings from two experiments. Archives of Sexual Behavior. https://doi.org/10.1007/s10508-025-03318-9