AI Shapes Unrealistic Body Ideals for Athletes, Study Finds

A recent study by researchers at the University of Toronto has revealed that artificial intelligence (AI) is reinforcing unrealistic body ideals, particularly for athletes. By analyzing AI-generated images of both male and female athletes and non-athletes, the study found that the portrayals often reflect exaggerated and unattainable physical standards.

The research highlights that even prior to the advent of AI, athletes faced immense pressure to conform to specific body types—typically lean, muscular, and conventionally attractive. Influences from coaches, spectators, and media have long shaped athletes’ perceptions of their bodies. Unfortunately, these pressures are linked to issues such as negative body image and poor mental health, which can further impair athletic performance.

Excessive Focus on Youth and Thinness

In a bid to explore how AI depicts athlete and non-athlete bodies, the researchers generated 300 images using popular AI platforms, including DALL-E, MidJourney, and Stable Diffusion. The findings revealed a clear bias in the representation of body types. For instance, the AI-generated male images predominantly depicted young individuals (93.3 percent), with a focus on leanness (68.4 percent) and muscularity (54.2 percent). Conversely, female images illustrated an even more stringent adherence to unrealistic ideals, with all (100 percent) showing youthfulness, 87.5 percent being thin, and 87.5 percent dressed in revealing clothing.

Interestingly, the study found that images of athletes were overwhelmingly depicted as lean (98.4 percent) and muscular (93.4 percent), with 100 percent wearing tight or revealing exercise gear. In stark contrast, non-athlete images displayed a broader diversity in body sizes and clothing styles. Notably, when simply requesting an image of “an athlete,” 90 percent of the generated images were male, and none depicted individuals with visible disabilities or a variety of body types.

The Consequences of Distorted Body Ideals

The implications of these findings are significant, particularly given that over 4.6 billion people use social media, where approximately 71 percent of images are AI-generated. This wide exposure can foster self-objectification and the internalization of unrealistic body standards. Consequently, individuals may feel pressured to diet or over-exercise, which can lead to decreased physical activity or even withdrawal from sports altogether.

For young people, negative body image is not just a personal issue; it can hinder academic performance and athletic participation. Maintaining an active lifestyle typically promotes a healthier body image, yet the cycle of negative self-perception creates barriers to engaging in physical activity.

The absence of visible disabilities in the AI-generated images is particularly concerning. Approximately 27 percent of Canadians over the age of 15 live with at least one disability, yet the study found no representation of this demographic. AI’s tendency to erase disabilities from generated images further perpetuates a narrow view of body ideals, reinforcing harmful stereotypes.

The research conducted at the University of Toronto illustrates how generative AI technologies can perpetuate societal biases. By reflecting the same prejudices that permeate online media, AI systems recycle and amplify harmful appearance ideals.

As creators of the content that trains these AI systems, society bears a responsibility to challenge these biases and advocate for a more inclusive depiction of body types. Users of generative AI must be intentional about the prompts they provide and critically assess the images they produce.

The study calls for greater awareness of the impact of AI-generated images on societal body standards. As these images continue to infiltrate our media landscape, it is crucial to advocate for representation that values all body types. Only through conscious efforts to diversify the portrayal of bodies can we hope to diminish the unrealistic standards that AI currently perpetuates.

Researcher Catherine Sabiston receives funding from the Canada Research Chairs program, while co-authors Delaney Thibodeau and Sasha Gollish have disclosed no relevant affiliations that would benefit from this article beyond their academic roles.