Abstract
Millions of people now use generative artificial intelligence (GenAI) tools in their daily lives for a variety of purposes, including sexual ones. This narrative literature review provides the first scoping overview of current research on generative AI use in the context of sexual health and behaviors.
Introduction
The rapid emergence of Generative AI (GenAI) tools—such as ChatGPT, Replika, and Midjourney—has fundamentally altered the landscape of digital interaction. While artificial intelligence has existed for decades, the recent shift toward generative models allows for human-like interaction in deeply personal and sensitive spaces. This paper synthesizes five years of research (2020–2024) to map how AI is transitioning from a niche "sextech" curiosity to a mainstream driver of sexual health, therapy, and digital intimacy.
The authors address a critical gap: the lack of a unified overview of the intersection between AI engineering and sexology. By identifying where AI offers opportunities, such as low-threshold access to health information, and where it creates genuine harm, such as AI-generated image-based sexual abuse (AI-IBSA), this review provides a foundational framework for future clinical and sociological study.
Methodology
The authors conducted a narrative scoping review, querying major databases including PubMed, Scopus, and Google Scholar. They utilized a combination of AI-related terms (e.g., LLM, chatbot, deepfake) and sexuality-related terms (e.g., contraception, pornography, intimacy). The search was restricted to peer-reviewed English publications from 2020 to mid-2024.
The final corpus consisted of 88 publications covering 106 individual studies. These were coded by study type (theoretical vs. empirical), research topic, and results, then categorized into four thematic clusters: education, therapy, companionship, and pornography.
Evidence Base: Evaluating GenAI Performance
A significant portion of the reviewed literature focuses on the accuracy and utility of GenAI in providing sexual health information. The tables below summarize 14 key studies that benchmarked AI performance against clinical standards.
Key Insights and Results
Information Quality and Clinical Utility
The review found that 12 out of 14 studies confirmed ChatGPT provides highly accurate and complete sexual health information, often outperforming traditional FAQ documents from organizations like the American College of Obstetricians and Gynecologists (ACOG). In specialized psychiatric evaluations of sexual dysfunction cases, ChatGPT’s responses received "A" or "B" grades 92% of the time.
The "Empathy Gap" and Arousal Bias
Surprisingly, several studies indicated that both laypeople and trained therapists rated AI-generated relationship advice as more empathetic and helpful than advice written by human professionals. However, there is a notable "anti-AI bias" in sexual arousal; experimental results suggest that heterosexual individuals report lower sexual arousal to erotic images when they believe the images are AI-generated, regardless of whether the images are real photos.
Anatomical Hallucinations and Deepfake Asymmetry
While LLMs excel at text-based clinical info, image generators (like DALL-E) frequently fail at accurate anatomical visualizations, often producing nonsensical or incorrect images of female anatomy. Furthermore, research is heavily skewed toward the harms of non-consensual deepfakes. Large-scale surveys indicate that 2% to 14% of the population has experienced some form of AI-generated image abuse, with women and minority groups being disproportionately targeted.
Limitations and Future Directions
The review identifies several critical gaps in current research:
- Lack of Longitudinal Data: The long-term psychological effects of forming "parasocial" romantic relationships with AI companions remain unknown.
- Demographic Bias: Most research focuses on male users of erotic chatbots, leaving the experiences of female users and sex workers under-researched.
- The "Black Box" of Moderation: Platform-level safety filters may inadvertently hinder access to valid sexual health information by blocking queries containing keywords like "breast cancer" or "sexual consent."
- Regulatory Lag: There is an urgent need for robust legal frameworks to address "synthetic" harm and AI-IBSA without stifling ethical innovation in the field.
As AI continues to integrate into the most intimate aspects of human life, ongoing interdisciplinary research between technologists and sexologists will be essential to navigate this evolving digital frontier.
Summary updated