Young Men Lead the Uptake of AI-Generated Pornography, Australian Survey Finds

Generative artificial intelligence has quietly rewritten the economics of explicit media. What once required cameras, performers, and production crews can now be conjured from a text prompt in seconds, and a new study suggests that audiences are adapting faster than many researchers expected. According to a survey of 367 heterosexual Australian adults published in the Archives of Sexual Behavior, roughly one in four respondents—26.0 percent—reported having used AI-generated pornography, a category that spans both fully synthetic imagery and deepfake content built on the likenesses of real people. The findings offer one of the first empirical portraits of who is actually engaging with this rapidly evolving medium, and the answer is strikingly specific: young men.

The research, conducted by Joel R. Anderson and Lily Moor of the Australian Research Centre in Sex, Health and Society at La Trobe University, arrives at a moment when the technology has outpaced the science surrounding it. AI-driven deepfake pornography first drew public attention around 2017, when online forums began circulating machine-learning-generated fake celebrity videos. The technology matured rapidly; by 2019, an estimated 96 percent of all deepfake videos on the internet were pornographic in nature. Early analyses warned that these hyper-realistic fabrications—which can make it appear that someone performed sexual acts they never did—pose serious threats to privacy and well-being, and deepfake porn has since been characterized as a new form of image-based sexual abuse that disproportionately targets women.

Yet the technology has a consensual side as well. A parallel market of uncensored generative AI tools and websites now allows users to create original explicit material on demand, tailoring body features, scenarios, and fantasy elements through prompting. This unprecedented interactivity and customization has fueled debate about whether AI-porn could reshape sexual media entirely—perhaps letting couples safely explore fantasies or serving groups underserved by mainstream porn, or perhaps normalizing fake imagery and distorting real-life sexual expectations. Notably, a recent review of 2020–2024 scholarship on AI and sexuality found that nearly all studies focused on non-consensual deepfakes, leaving a glaring gap: almost nothing was known about who engages with AI-generated sexual content in the broader sense.

To fill that gap, Anderson and Moor surveyed Australian adults aged 19 to 87 (mean age 39.07) between August and September of 2025. The sample comprised 52.5 percent men, 44.8 percent women, and a small number of non-binary participants who were excluded from inferential analyses due to cell sizes. Participants were recruited online and required to be living in Australia, heterosexual, and at least 18 years old; the restriction to heterosexual participants reflected a deliberate decision that sexual minority populations warrant independent study. After excluding eight participants who failed an attention check, the researchers analyzed responses to two key outcomes: a five-point familiarity scale and a yes-or-no question about personal experience using AI-porn, defined for participants as sexually explicit content created with artificial intelligence, including deepfakes and self-generated material.

The statistical architecture of the study was deliberately exploratory. Because AI-generated porn differs from conventional porn in its personalization, synthetic realism, and deepfake production, the authors declined to make directional hypotheses. Instead, they examined three clusters of potential correlates drawn from the broader pornography literature: demographic factors (age and gender), ideological factors (religiosity and political orientation), and relational factors (relationship status, relationship satisfaction, and sexual satisfaction). Analyses combined bivariate correlations with forced-entry multiple linear regression for familiarity and binary logistic regression for use, with bootstrapping across 1,000 samples. A sensitivity power analysis indicated the study could detect small-to-moderate effects, though power was more limited for predicting use, since only 94 participants reported having engaged with AI-porn.

The demographic signal was unambiguous. Men reported significantly greater familiarity with AI-porn than women (mean 2.91 versus 2.06 on the five-point scale), a large effect corresponding to Cohen’s d of 0.82. Gender differences in actual use were also substantial, with a Cramér’s V of 0.35. In the regression models, gender and age were the only significant predictors of either outcome. The logistic regression revealed that men’s odds of having used AI-porn were approximately 8.75 times higher than women’s, while each additional year of age slightly reduced the likelihood of use (odds ratio 0.96). The familiarity model accounted for 21 percent of the variance, and the use model explained between 18.2 and 26.6 percent of variance while correctly classifying 75.3 percent of cases. About 9.1 percent of women in the sample reported AI-porn use, indicating that women are engaging with the medium, albeit at much lower rates.

What is arguably more revealing is what did not predict engagement. Religiosity, political orientation, relationship status, relationship satisfaction, and sexual satisfaction were all unrelated to both familiarity and use. This null result defies expectations from the conventional pornography literature, in which religious and politically conservative individuals consistently report lower use and more negative attitudes. The authors caution, however, that their sample was only moderately religious on average and clustered near the political midpoint, which may have restricted variability and blunted sensitivity to ideological effects. Still, the pattern raises an intriguing possibility: the novelty of AI-generated content may appeal across usual ideological lines, drawing in people who might abstain from conventional porn for moral reasons.

The relational null findings carry their own weight. Popular narratives often frame pornography as either a cause or a consequence of relationship difficulty—the unhappy partner turning to explicit media as a substitute outlet. But in this dataset, AI-porn use was not associated with being single or partnered, nor with how satisfied people were with their relationships or sex lives. This aligns with a broader body of research showing that porn use occurs across diverse relationship contexts and is not inherently indicative of relational strain. The authors suggest that early AI-porn engagement may reflect curiosity, recreation, or experimentation with a new technology rather than compensatory need—many users may have tried it only a few times, intrigued by the tool itself rather than using it as a habitual coping mechanism.

Situated in the longer history of media adoption, the findings look less surprising. The adult entertainment industry has historically driven uptake of new technologies, from VHS tapes to online streaming and credit card systems, and the current AI-porn landscape—shaped by male-dominated deepfake forums and custom generator sites—likely reinforces male-centric engagement. The well-documented gender gap in conventional porn use, attributed to a mix of biological, social, and cultural factors, appears to carry over intact into the AI era, while younger adults’ greater familiarity mirrors patterns in digital innovation generally. Whether older adults will follow as the technology becomes more accessible remains an open empirical question.

The study’s limitations are considerable and the authors are candid about them. The sample was predominantly heterosexual and cisgender, leaving LGBTQIA+ populations unexamined. The broad use measure did not distinguish between fully synthetic content, deepfakes of real people, or AI-generated text and chat material—categories with vastly different ethical and legal implications—nor between deliberate seeking, incidental exposure, viewing others’ creations, and generating content oneself. The study was not pre-registered, no multiple-testing corrections were applied, and the many statistical tests elevate Type I error risk, though the consistency of the age and gender effects across outcomes offers some reassurance. Crucially, the study identifies who is engaging with AI-porn, not how engagement affects users; whether hyper-personalized synthetic media will heighten unrealistic expectations, blur consent norms, or remain a largely benign curiosity is a question for longitudinal research. For clinicians and educators, the practical message is already clear: with a quarter of adults reporting use, practitioners should expect that some clients are encountering AI-generated sexual media and should ask about it in routine, non-stigmatizing ways rather than assuming a particular moral, political, or relational profile.

Subject of Research: Demographic, ideological, and relational correlates of AI-generated pornography use and familiarity among Australian adults

Article Title: Who Is Using Pornography Generated by Artificial Intelligence? Correlates of AI-Porn Use and Familiarity in an Australian Adult Sample

Article References: Anderson, J. R., & Moor, L. (2026). Who Is Using Pornography Generated by Artificial Intelligence? Correlates of AI-Porn Use and Familiarity in an Australian Adult Sample. Archives of Sexual Behavior. https://doi.org/10.1007/s10508-026-03584-1

Image Credits: AI Generated

DOI: 10.1007/s10508-026-03584-1

Keywords: artificial intelligence, pornography, deepfakes, digital sexuality, gender differences, sexual behavior, survey research, Australia, relationship satisfaction, religiosity, generative AI, sex research