Recent research published in the Archives of Sexual Behavior reveals that artificial intelligence can generate erotic images of women that viewers find more aesthetically pleasing and sexually attractive than actual photographs of human beings. While people still recognize actual photographs as looking more authentic, the fabricated images reliably score higher in measures of attractiveness and overall pleasantness. These results suggest a notable shift in how digital technology might alter human perceptions of physical beauty and sexual appeal.
Lead researcher Ellen Zakreski, a scientist at the Czech National Institute of Mental Health and Charles University, collaborated with colleagues to investigate modern perceptions of visual sexual material. The research team pursued this inquiry because expectations surrounding erotic imagery have shifted across different generations. Images that researchers validated for psychological testing several decades ago often fail to elicit the same reactions from younger adults today.
Humans have created visual sexual materials for thousands of years. Early examples range from ancient cave drawings and fertility statues to vintage pornographic magazines. Today, the advent of the internet has caused an explosion in the availability and variety of digitized sexual media.
Viewers can now instantly access relatively realistic depictions of sexual acts as well as highly stylized, fantasy-laden content. This modern media landscape includes artificial intelligence and computer-generated imagery. These digital tools allow creators to build highly realistic, three-dimensional images of people who do not actually exist.
While traditional plastic surgery can alter the appearance of real humans, digital programs give creators absolute control over every physical detail. Artificial intelligence, in particular, can generate ultra-realistic images with incredibly sharp visual resolution. This technology can alter a real photograph by changing a person’s body shape, facial features, or hair color.
It can also fabricate entirely new people from scratch. Additionally, modern viewers frequently encounter images of real women who have undergone surgery to enlarge their lips, breasts, or hips. Other distinct categories of sexual media include lifelike silicone sex dolls and hentai.
Hentai is a pornographic form of Japanese comic art that features highly exaggerated physical proportions and fantasy elements. With super-realistic sexual imagery becoming widely accessible, viewers may struggle to distinguish between real photographs and digital creations. Zakreski and her colleagues wanted to understand how people respond to these different visual styles.
To explore these modern visual categories, the researchers conducted a nationwide online survey in the Czech Republic. They recruited 649 adults who self-identified as being sexually attracted to women. The participant pool consisted mostly of men, though a smaller group of 45 women also completed the survey.
During the experiment, participants viewed static images of naked female figures presented against a neutral gray background. The research team exposed the volunteers to six distinct categories of images. These categories included actual photographs of real women, computer-generated figures, figures created by artificial intelligence, real women with surgically enhanced features, silicone sex dolls, and hentai illustrations.
To account for different personal preferences, the researchers ensured that each of the six categories contained five distinct character types. These figures varied by hair color, ranging from blond to black. They also featured different body types, such as voluptuous, athletic, and petite.
The team purchased basic three-dimensional digital templates and refined them to create the computer-generated figures. They then searched for real photographs, sex doll images, and hentai illustrations that closely matched the poses and body types of those digital models. They also used an online artificial intelligence platform to generate matching figures through specific text prompts.
After selecting the images, the team used photo editing software to standardize the visuals. They removed distracting elements like tattoos and jewelry. They also matched the skin tones across all categories to look like the original computer-generated figures, ensuring a level of visual consistency.
Participants rated each image using a digital slider scale from zero to 100 for realism, sexual attraction, and aesthetic appeal. They also rated the emotional pleasantness of each image, a concept scientists call valence, using a standard five-point pictorial scale. The researchers randomized the order of the images to prevent any sequential biases.
This pictorial scale used simple cartoon figures to measure emotional responses. It helped participants indicate whether an image made them feel “unhappy, annoyed, dissatisfied, melancholic, desperate, or bored” or whether it made them feel “happy, pleased, satisfied, hopeful, or relaxed”. Participants selected the cartoon figure that best matched their emotional reaction.
When reviewing the data, the research team found that participants rated the actual photographs as the most realistic category. Artificial intelligence images took second place in terms of realism. The computer-generated figures, surgically enhanced women, sex dolls, and hentai illustrations followed in descending order of perceived realism.
Despite being seen as less realistic than actual photographs, the images generated by artificial intelligence ranked the highest in aesthetic appeal. Participants also rated the artificial intelligence figures as the most sexually attractive and the most emotionally pleasant. Actual photographs ranked second across these three metrics.
The images of women with surgically enhanced features received the lowest ratings for aesthetic appeal and overall pleasantness. Hentai illustrations and images of sex dolls also ranked low in these categories. Most participants found these less realistic categories to be somewhat unpleasant overall.
The study revealed distinct differences in how men and women evaluated the images. Men generally gave higher ratings across all categories for realism, attractiveness, aesthetic appeal, and pleasantness. Women found the images less appealing and less realistic overall, though their general ranking of the categories matched the pattern seen among the men.
Age also played a major role in how participants perceived the different categories. Older individuals tended to give higher aesthetic and sexual attractiveness ratings to the actual photographs, artificial intelligence images, and computer-generated figures. They also found these realistic categories more emotionally pleasant than younger participants did.
Conversely, younger participants responded much more favorably to the hentai illustrations. They rated these stylized cartoons as more aesthetically pleasing, sexually attractive, and emotionally pleasant than older participants did. The researchers suggest that younger generations have greater exposure to Japanese comic art styles in mainstream media, which might normalize the aesthetic for them.
The researchers note that these age-related differences highlight how early exposure to certain types of media might shape long-term visual preferences. Older adults likely grew up viewing photographs of natural women in magazines. Younger adults have grown up in an era saturated with digital manipulation and highly stylized animations.
The research team acknowledged several limitations in their experimental design. The participant sample included far more men than women, which makes it difficult to draw broad conclusions about female preferences. Additionally, the recruitment methods targeted individuals already registered in a sexuality research database.
This database included people engaged in outreach programs for atypical sexual interests, meaning the sample might not perfectly reflect the general public. Another limitation involves the physical characteristics of the images used in the survey. All the female figures featured lighter skin tones to align with the general ethnic demographics of the Czech Republic.
The researchers also had to rely on preexisting images for several categories. This limitation resulted in minor inconsistencies, as some figures had different facial expressions, hand placements, or hairstyles. The study also relied entirely on subjective self-reporting rather than objective physiological measurements.
The researchers did not track physical signs of sexual arousal, such as heart rate or blood flow. Future studies could incorporate these physiological metrics to see if the human body reacts differently to real photographs versus digital creations. The authors recommend that future investigations expand the variety of the visual stimuli.
They suggest creating image sets that feature diverse ethnic backgrounds and different age ranges. Expanding the research to include images of naked men would also help clarify whether these preferences for artificial intelligence imagery hold true for people attracted to males. As digital technology continues to advance, artificial intelligence will likely produce even more convincing and stylized erotic media.
The study, “Subjective Responses of Gynephilic Men and Women to Real versus Artificial Female Nudes,” was authored by Ellen Zakreski, Alena Marečková, Ondřej Vaníček, Martin Hůla, Kateřina Klapilová, Jitka Lindová, and James G. Pfaus.
Leave a comment
You must be logged in to post a comment.