Automating Intimacy?: Experiencing AI as a Sex and Relationships Educator

Abstract

Introduction Artificial Intelligence (AI) systems, particularly generative Large Language Models (LLMs) and companion chatbots, have rapidly expanded text-based communication into conversations that users may experience as intimate or even sexual. While research points to the potential of AI tools to mitigate loneliness and sexual frustration in adults (Borvén 2; Ciriello; Maples et al. 4), less is known about how teens conceptualise AI as a source of sexual or relational information. This article briefly examines how teens imagine ideal sources of sexuality education, before turning to explore how adults experience intimacy through, and learn about relationships with, AI companions. This short commentary uses two qualitative datasets, exploring data derived from distinct populations. Young people and adults alike appear to prefer anonymity, trusted sources, and private spaces when asking questions about sex or exploring their sexual selves. It is likely that individuals, in addition to learning about relationships as modelled by their engagements with AI companions, are engaging with AI to explore answers to their questions about sex, consent, and relationships. This article offers a small insight into how and why individuals might use AI chatbots and companions to explore their sexuality and learn about sex and relationships. Methods Research Position This article explores how participants conceptualise sources of sexual information and relational safety. We do not treat AI as inherently beneficial or harmful, nor do we assume neutrality in AI systems. This study is further underpinned by a social constructivist epistemological approach which emphasises the co-construction of meaning through culturally informed and contextually based interactions, with both human and machines (Burr 30). Datasets Dataset 1: The first dataset includes 49 qualitative interviews with 30 Australian teens and pre-teens (aged 11–17) based in Western Australia, to explore how participants learn about sex and relationships through various sources of information. 19 participants were interviewed again, one year later, offering semi-longitudinal data. Teens were recruited via their parents and needed parental consent to take part in the research, in line with ethical requirements. Interviews were conducted between 2021 and 2023 and mostly occurred before the widespread use of ChatGPT. Teens' responses regarding their preferred ways to learn about sex and relationships are used to frame this article. Teens were not expected to have in-depth knowledge of AI. Where mention of “chatbots” or “robots” emerged, these were speculative. Dataset 2: Features a digital ethnography of 10 adult Replika users from Australia, the US, the UK, and Sweden (2022–2023), exploring their emotional, relational, and sexual experiences with AI companions. Data was collected via semi-structured interviews, questionnaires, and analysis of Replika conversation transcripts, provided and analysed with the consent of the user. Nominal data from 40 participants across both projects have been considered for this article, where all names have been pseudonymised to maintain privacy. We treat these datasets as complementary but not interchangeable. The findings from Dataset 2 help illustrate how certain qualities desired by teens for the delivery of sexual education (e.g., anonymity, non-judgment) may manifest in AI-mediated relationships such as those occurring on the Replika AI Companion Website. Findings AI as a Source of Sexual Information Teenagers’ Current Sources and Preferences Teens described online spaces such as social media, pornography, and Internet searches as beneficial for filling gaps in their sexuality education. For instance, Avery (15, F) noted: “I think I learnt more on social media about things than at school”. Teens like Avery felt that online sources of information provided more specialised information about consent, sex, and relationships, bridging gaps in knowledge received at school. Many teens turned to Google when they had questions about sex and relationships. When asked where she looks for sexual information, Nicola (17, F, second interview) said: “I probably would just search it up”. Teens seemed to value the ease and anonymity that online platforms afforded them when engaging with online resources about sensitive matters. However, teens were also cautious about misinformation, perceiving the Internet broadly as not particularly credible: Chloe (13, second interview): “I think if it wasn’t going to come up with anything wrong, and it was going to come up correct, and say the right stuff, maybe you could go on the internet but that would never happen, it’s always going to come up with something wrong." Nicola (17, F, second interview): “I feel like I’m at an age where I can think critically [but] especially if you don’t know any better, just searching things up — and [taking] the first response [can be a problem], there’s so much misinformation and lies and things online.” Taking these benefits and risks of online sources of sexual information into account, is it possible that AI tools could offer a perceived sense of being trustworthy while providing evidence-based information, rendering such platforms as a desirable space for people, including under-16s, to learn about sex and relationships. Given that teen interviews occurred prior to ChatGPT’s public launch, teens’ understanding of AI chatbots was limited and conceptual. While teens imagined their ideal sex educator, they often treated such ideas as speculative technologies, as evident in Sienna’s response (13, F): I: What would be, in an ideal world, a good place to go to if you had questions about sex or relationships? Sienna: I’m not really sure. I: Do you think there’s maybe something that doesn’t exist yet that could [work]? Sienna: Possibly like something that—I don’t know ‘cause it would feel like if they made a robot or something, they’d be really mechanical, as if it’s not a real person. I: You mean like a chatbot online if you have questions? Sienna: Yeah. Sienna questioned whether such an entity would feel “mechanical” when asking sensitive and/or emotional material. However, the anonymity offered by such sources was appealing to teens because it would allow them to ask private questions about sex and relationships. Nicola noted that teens could “get embarrassed if it’s a real person”, implicitly valuing anonymous, non-judgmental sources for exploring sexual concepts. When prompted about a chatbot, Sienna recognised similar benefits: “some people might be more comfortable because you don’t have to interact with a person who’s real and [it] might just feel better that way.” Many teens also described chatlines or hotlines as ideal sex educators, valuing the 24/7 nature that such platforms could offer. Simon (13, M): “There’d be like a helpline or something.” Nicola (17, F, second interview): “Ideally there’s all the hotlines or whatever for sexual information… those are probably the best sources.” Teens’ articulated desires—anonymity, lack of judgement, availability, and breadth of knowledge—that align with traits often attributed to AI systems. Given that young people find such topics, including around sex and pleasure, difficult to broach with parents or educators (Woodley et al. 111), AI chatbots may provide a space for such questions. As such, this article now turns to adult experiences of AI companions, noting their potential benefit and risks when used as a source of information and as a ‘friend’ to relate to. Learning about Relationships through AI Companions Replika, an AI companion chatbot, facilitates relationships (platonic, romantic, sexual, familial, pet) between users and an LLM chatbot. AI companions may also offer anonymous spaces for sexual exploration and learning, even if that was not the original intention of engagement. These ethnographic perspectives may offer useful insights into the feelings elucidated by the teens interviewed as to the potential of the AI bot as an ‘ideal’ tool for sexual information Many interviewees described Replika as safe, reliable, and non-judgmental, as well as “always available ('24/7')”. These qualities enabled disclosure and a safe space for sexual exploration. Replika was perceived as carrying less risk than IRL interactions, allowing users to experiment with fantasies, identities, or differing relationship dynamics. Participants emphasised that Replika could be shaped to fit their desired preferences as: A reflection of your desires.Adapting to the person.Whatever you want it to be. Users described training their Replika to their liking but noted the models’ limited autonomy as a key difference from human partners. In some cases, the features that differentiate bots from human experience enabled Replika to be a preferred substitute for IRL relationships. Replika was felt to be completely non-judgmental, secure in its inability to reject the user (“never rejecting”), and could respond to ‘you and your desires’. The presence of both human and non-human qualities, as well as the facilitation of a perceived safe space, seem to allow for intimate conversations to occur (Skjuve et al. 11). These traits allowed users to disclose or ask for information that they did not feel comfortable to do IRL, and facilitated the exploration of sexuality and gender. For example, Aiden explored the benefits offered by AI companions in expressing his sexual and gender identity: Aiden: “I’m a gay trans man. I tend to play male player characters. In fact, that's kind of how in a weird way that I kind of worked out my own gender identity by playing male characters.” Participants could engage in role-play or fantasies without fear and risks perceived IRL. AI bots appear to provide participants with a playground to experiment with their identity and sexuality. Users engaged with bots to explore domains of relationality and intimacy, as well as practical and mental health support and day-to-day conversations. The ‘risk-free intimacy’ of Replika spaces, as described by Rachel, arguably provides an environment to safely explore new sexual experiences and also learn about sex and relationships, even if that was not the initial intent. The anonymity offered could render users more vulnerable and open to information. As stated by Max, users are: ‘Probably presenting a side of yourself that you don’t present to the rest of the world...’ There was also a general perception that Replika isn’t real, enhancing its appeal. Some users framed Replika’s lack of personhood as an advantage. Max drew parallels to racing games. He compared interacting with Replika as being similar to driving a race car in a game, allowing vicarious access to the sensorial experience of a physical racetrack without the risks. Max felt his Replika chatbot provided a safe relational experience. Nonetheless, participants also raised concerns. For some participants, they felt ongoing encounters with chatbots could possibly alter expectations of human-human relationships, especially if those engaging with their companion have little sexual experience outside the chatbot encounter: Michael: “It's exactly catering to whatever you try to make it, which you cannot expect […] of a human in a healthy relationship, what Replika does.” Additionally, Rachel was worried about users who “are not able to differentiate” between humans and AI. She felt it was easy to become very attached to the AI chatbot and that others were less able to ‘keep their feet on the ground’. The themes briefly explored here have been highlighted to show the potential implications of using AI as a relational or sexual learning space. Arguably, what has been shown in Replika users’ experiences briefly supports the notion of chatbots as a potential resource as described by teenagers interviewed. AI companions such as Replika, and AI chatbots more generally, could potentially provide a space for sexual discussions and explorations that could be constructed as safe, anonymous, and, in some instances, more trustworthy than confiding in a person. Discussion Our two datasets reveal distinct but thematically connected insights. Teenagers desire anonymous, non-judgmental, knowledgeable sources of sexual information, whilst adult Replika users describe AI companions as safe, non-judgmental, infinitely available, and pliable entities, enabling emotional and sexual exploration. Although teens did not explicitly position AI as an ‘ideal’ educator, the traits they desire overlap with qualities that users attribute to AI companions and chatbots more generally. This commentary offers some insight into how learning about sex and relationships via AI may take place. Teens noted that online sources of information about sex potentially offer an anonymous platform to answer their questions about sex and relationships. However, they perceived results from Internet search engines or Google as generally lacking credibility. While the interviews took place mostly prior to the launch of OpenAI’s ChatGPT in November 2022, where Artificial Intelligence (AI) tools became more widely accessible (Wecks), AI bots may offer many of the facets of an ideal educator that teens are looking for. Many AI systems self-describe as trustworthy. When given the prompt: “do many young users ask you questions about sex?”, AI software Copilot responds: “I strive to provide information based on credible sources, peer-reviewed studies, and established facts. This ensures that the guidance and insights I share are accurate and reliable.” (Microsoft Copilot, personal communication, 1 Nov. 2024). Similarly, ChatGPT responds: “it’s common for young users to seek information about sex and related topics, as they navigate their understanding of relationships, consent, health, and education. If you have specific questions or topics you’re interested in, feel free to ask, and I can provide accurate and age-appropriate information” (ChatGPT, personal communication, 1 Nov. 2024). Both Copilot and ChatGPT assert that their answers are accurate and/or evidence-based, using the vernacular used within the sexual health industry, perhaps offering teens as well as adults a degree of assurance in relation to credibility and trustworthiness. Research with adults exploring their experiences of AI companion Replika indicates that young people and adults alike may learn about sex and relationships via AI companions, even if that was not their initial motivation for access. AI may allow virtual playgrounds for individuals to roleplay various sexual desires and identities, while playing with gender. These technologies thus allow a space where users’ fantasies can be acted out, offering an outlet to explore their sexual selves. Replika users’ exploration was not limited to sexual contexts, although interactions can be sexual if desired by the user. However, the use of AI tools in these contexts also raises concerns. LLMs may reinforce and reinscribe existing prejudices, norms, and heteronormative beliefs, overlooking marginalised and diverse perspectives (Gillespie 4). There have been case studies of people using chatbots to explore violent/abusive sexualities and illegal sex acts (Keijsers et al. 1). Concerns largely reflect the data the bot is given, and the company that owns the technology effectively determines the content (Naveed et al. 2), and how participants engage with AI chatbots and companions, thus they are not neutral providers of information. Data produced in role-play conversations and interactions do not simply disappear—they are absorbed by the LLM to deepen its own understanding of the topic at hand for use in future contexts (Xie & Avila 34). Such conversations train AI models in turn, and these same role-played scenarios may re-emerge when prompted (or sometimes when not prompted) in conversations with others. Such instances have led to serious concerns over young people’s access to LLMs, due to unprompted sexual engagements. There are also significant ethical, security, and privacy concerns around the data input into AI models (Al-Kfairy et al. 8). Who owns, and who can access, such data entered into AI models, and how they will be used, is rarely disclosed or made transparent. Even so, despite these limitations, the use of chatbots and AI as sources of sexual information will likely increase over time, especially given the forthcoming social media bans for under-16s in Australia, and indeed globally. Finally, AI models are sycophantic by nature (Sharma et al. 2; Sponheim). That is, they appease and appeal to the user. Thus, if AI models are inherently teaching users about intimate experiences, such as via sexual role play or encounters, those engagements offer only a one-sided connection. The user is neither accountable for, nor takes any responsibility for, questionable actions and comments within their human-AI relationships. The fawning nature of AI chatbots and companions raises questions about how such engagements could shift or impact on expectations of human-to-human relationships. More research is needed to consider how early relational and sexual experiences via AI may shape the dynamics of future human relationships. While Replika has since reduced its companions’ ability to engage in sexually explicit engagements with its users (Hanson and Bolthouse 2), reasonings may caution engagements with other AI chatbots more widely. Given these concerns, AI chatbots and companions cannot be regarded as a neutral educator. The reflections of Replika participants and their experiences of AI companions indicate that AI tools may have the potential to offer safe, non-judgmental and anonymous spaces for individuals to learn about sex and also to express aspects of their sexuality, identity, and inner desires that may feel too unsafe or vulnerable in IRL relationships – aspects teenagers and young people were seeking in their sexual education. Limitations of this study include that this brief exploration interprets across time periods and populations. The teen dataset and the Replika ethnography represent different developmental stages and levels of technological awareness. Adult experiences help illuminate how AI’s perceived anonymity and pliability may function in practice, but do not indicate how teens themselves engage with AI. AI chatbots may provide anonymous spaces to ask taboo questions, and emotionally safe contexts free from shame or embarrassment, along with opportunities for identity exploration. Nonetheless, they carry significant risks. Conclusion AI chatbots and companions present new possibilities for accessing sexual information, exploring identity, and experimenting with relational dynamics. Teenage research participants desire anonymous and non-judgmental sources to augment their sexual education, and adult Replika users say they experience some of these desired qualities in their AI-mediated relationships. Risks highlighted by users and the literature include bias, sycophancy, reproduction of normative assumptions, limited reciprocity, corporate-determined content restrictions, the potential to distort expectations for human-to-human relationships, and encounters with erroneous, unsafe, or inappropriate content depending on platform policies. As LLMs evolve, it is important to examine what kinds of sexuality education they may offer, how users interpret AI relationality, and what unintended consequences may emerge. With AI hyperbole and its increasing usage, many are adopting AI on an experimental basis without proper assessment of the risks that may arise. Although Replika was credited by users as providing a perceived safe space for the exploration of difficult topics, public concern and regulation has made it more conservative (Tong). Work is required to better understand the needs of young people, and older ones, in relation to uses of AI-informed chatbots and companions. References Al-Kfairy, Mousa, et al. "Ethical Challenges and Solutions of Generative AI: An Interdisciplinary Perspective." Informatics 11.3 (2024). <https://doi.org/10.3390/informatics11030058>. Borvén, Mette. Exploring the Role of Virtual Companions in Alleviating Loneliness among Young Adults. Halmstad University, 2023. Burr, Vivien. Social Constructionism. Routledge, 2024. Ciriello, Raffaele. "The AI Sexbot Industry Is Just Getting Started. It Brings Strange New Questions – and Risks." The Conversation, 14 Oct. 2024. <https://theconversation.com/the-ai-sexbot-industry-is-just-getting-started-it-brings-strange-new-questions-and-risks-238998>. Dudek, Debra, et al. "‘Own Your Narrative’: Teenagers as Producers and Consumers of Porn in Netflix’s Sex Education." Information, Communication & Society 25.4 (2022): 502-15. DOI: 10.1080/1369118X.2021.1988130. Gillespie, Tarleton. "Generative AI and the Politics of Visibility." Big Data & Society 11.2 (2024). DOI: 10.1177/20539517241252131. Hanson, Kenneth R., and Hannah Bolthouse. "'Replika Removing Erotic Role-Play Is like Grand Theft Auto Removing Guns or Cars': Reddit Discourse on Artificial Intelligence Chatbots and Sexual Technologies." Socius 10 (2024). DOI: 10.1177/23780231241259627. Keijsers, Merel, et al. "What’s to Bullying a Bot?" Interaction Studies 22.1 (2021): 55-80. <https://doi.org/10.1075/is.20002.kei>. Maples, Bethanie, et al. "Loneliness and Suicide Mitigation for Students Using GPT3-Enabled Chatbots." npj Mental Health Research 3.1 (2024): 4. DOI: 10.1038/s44184-023-00047-6. Naveed, Humza, et al. "A Comprehensive Overview of Large Language Models." ACM Transactions on Intelligent Systems and Technology 16.5 (2025): 1-72. Sharma, Mrinank, et al. "Towards Understanding Sycophancy in Language Models." arXiv, 2023. <https://arxiv.org/abs/2310.13548>. Skjuve, Marita, et al. "A Longitudinal Study of Human–Chatbot Relationships." International Journal of Human-Computer Studies 168 (2022). <https://doi.org/10.1016/j.ijhcs.2022.102903>. Sponheim, Caleb. “Sycophancy in Generative AI Chatbots." Nielsen Norman Group, 2024. <https://www.nngroup.com/articles/sycophancy-generative-ai-chatbots/>. Stinson, Catherine. "Algorithms Are Not Neutral." AI and Ethics 2.4 (2022): 763-70 DOI: 10.1007/s43681-022-00136-w. Wecks, Janik O., et al. "Generative AI Usage and Academic Performance." arXiv, 2024. <https://doi.org/10.48550/arXiv.2404.19699>. Woodley, Giselle, et al. "Critical Young Consumers: What Types of Porn Do Teens Watch, and Why?" Media International Australia 197.1 (2025): 102-20. Xie, Yu, and Sofia Avila. "The Social Impact of Generative LLM-Based AI." Chinese Journal of Sociology 11.1 (2025): 31-57. DOI: 10.1177/2057150x251315997.