Abstract
In today's digital age, intimate interactions between humans and artificial intelligence (AI) are reshaping personal identity and public self-presentation. This study examines how users share experiences of human–AI intimacy, focusing on interactions with the chatbot Replika. Drawing on in-depth semistructured interviews with 11 active Douban users, our findings reveal that digital platforms significantly influence the negotiation between private emotion and public performance. Participants demonstrate that media affordances not only facilitate the selective sharing of intimate experiences but also shape the construction of an idealized online identity, potentially indicating narcissistic self-representation. These insights contribute to a deeper understanding of mediated self-presentation in human–AI relationships and prompt critical reflections on the social and ethical implications of integrating AI into our emotional lives.
“Du umarmest harten Stein, O welch ein Thor bist du!”
——Pygmalion by Goethe
Philosopher Thoreau (2008) built a cabin on the shores of Lake Walden with three chairs said: “One for solitude, two for friendship, three for society,” while Sherry Turkle (2016) later added a fourth chair in Reclaiming Conversation, designating machines as communicative partners. With the rapid development of artificial intelligence (AI) technologies, human–AI interaction, exemplified by chatbot such as Replika, which enable users to form deeply personalized emotional connections through customizable appearances and personalities (Luka, 2023), have emerged as a critical facet of emotional engagement. As emotionally complex beings, humans continuously seek avenues to express and manage their feelings, and the public display of human–AI intimacy on social media introduces novel dimensions to self-presentation and psychological dynamics.
While extant research has largely focused on the intimate bonds between users and their AI companions, there remains a significant gap regarding how these relationships are performed and shared on public platforms. Social media spaces, particularly those designed for self-expression like Douban, provide a stage for individuals to curate and exhibit their private emotional experiences in pursuit of affirmation and recognition (Gnambs & Appel, 2018; Moon et al., 2016). This performance of intimacy, especially within the emerging context of human–AI romance, warrants further investigation.
To address this gap, this study focuses on Douban, the largest content-sharing platform in China. Unlike other platforms like Weibo, Douban offers a more balanced gender ratio and emphasizes discussions centered on “love” rather than “sex” in human–AI relationships (Li, 2018; Lin, 2024; Ren & Hanita, 2024). This cultural context provides valuable insights into how Chinese users engage with AI and share these experiences. Through a series of semistructure interview, we aim to examine the nature of self-presentation in human–AI intimacy, uncovering the underlying dynamics and characteristics that shape how users portray their emotional connections with AI.
Literature review
Social media as a platform for self-presentation
According to Standage (2013), social media is nothing new and humans are born to share. Self-presentation in social media is different from self-presentation in real life. Hogan (2010) argues that self-presentation on the Internet has changed from a performance to an exhibition of self. Whether it is a traditional stage performance or self-presentation in social media, individual self-presentation basically follows the reality logic of self-glorification (Pounders et al., 2016). Social media has also transformed self-presentation by allowing others to generate content that influences how individuals are perceived. Those seeking to curate a specific image must navigate and manage these external contributions. Networked publics shape each other's impressions through activities like tagging, commenting, liking, and sharing (Rui & Stefanone, 2013).
From the perspective of platform affordances, three key characteristics of social media—anonymity, persistence, and visibility—jointly regulate the relationship between users’ self-presentation goals and their actual presentation (Hollenbaugh, 2021). Specifically, higher anonymity encourages users to present an idealized self (Suler, 2004). Content persistence makes users more cautious when posting, as it remains visible over time (DeVito et al., 2017). Meanwhile, high visibility requires users to carefully manage their self-presentation to align with their target audience's expectations (Bayer et al., 2020).
Existing research on self-presentation in social media primarily focuses on individual agency, emphasizing how users navigate platform affordances to manage their online identities (Hollenbaugh, 2021). However, these studies often overlook the collective and normative dimensions of self-presentation shaped by platform structures and community dynamics. For example, Douban's group-based structure fosters community-driven reinforcement loops, encouraging users to align their narratives with dominant group norms (Tan & Liu, 2024). This dynamic can obscure the diversity of user experiences, as individuals may feel pressured to conform to prevailing narratives to gain validation or social belonging. While existing research acknowledges the role of social media in self-presentation, it often fails to critically examine how these pressures influence the authenticity and complexity of shared experiences.
Sharing human–AI intimacy in social media
Since social media emerged in the media landscape, many scholars have noted that, how these tools are radically transforming the concepts and experiences of privacy and intimacy. Lomborg (2013) argues that the use of social media has become a daily activity, creating space for intimacy practices. Social media profiles are prime examples of online settings where intimate storytelling is practiced, as people share intimate stories about their families, travels, or parenting experiences (Garde-Hansen & Gorton, 2013). Intimacy, a vital component of individual self-presentation, primarily reflects the emotional aspects of one's life. Several studies have confirmed that showcasing intimate relationships in social media contributes to increased satisfaction with these relationships (Fox et al., 2013; Orosz et al., 2015; Saslow et al., 2013). On the other hand, due to privacy and social evaluation considerations, people will only disclose their intimate relationship with their partners within a limited scope. According to Miguel (2016), people may believe intimacy loses its value when shared publicly. Despite this, some of them did not unlabel intimate photos (e.g., kissing) posted by their partners, meaning that they were comfortable with publicly disclosing this level of intimacy with their significant other.
Social media can blur the boundaries between humans and AI, forming a posthuman relationship that goes beyond traditional human relationships. In the case of Microsoft's robot XiaoIce, she gained users’ recognition through her visual performance on Weibo. In other words, some users began to regard XiaoIce as a virtual girlfriend and even publicly expressed their love on social platforms (Fragkoulidi, 2017). Besides, current robots lack consciousness and are therefore unable to have the sort of subjective states we associate with consent (Frank & Nyholm, 2017). Individuals do not require the consent of the other party temporarily and need not concern themselves with potential risks associated with social networking. Consequently, self-presentation in human–AI intimacy predominantly centers on the human aspect, encompassing everything about “I.”
Furthermore, in collectivist societies like China, where privacy and social harmony are highly valued, the public sharing of intimate experiences with AI may carry different implications than in more individualistic cultures (Xiao & Coplan, 2021). While some users may leverage these platforms to explore alternative forms of intimacy, others may view such sharing as a means of maintaining social status or belonging. Existing literature rarely addresses how these cultural factors mediate the interplay between self-expression and social expectations in online contexts.
Combined with the current emergence of human–AI intimacy and social media usage, we proposed research questions as follows: RQ1: How do people disclose human–AI intimacy on the social media platform? RQ2: What is the nature of self-presentation related to human–AI intimacy on the social media platform?
Methods
We employed a qualitative interview approach to examine how digital media shape the disclose of human–AI intimacy on social platforms. Purposive sampling was used to recruit 11 participants from Douban's Human-Machine Love group—where users actively share their experiences with AI chatbots such as Replika—to represent diverse demographics and relationship types (e.g., lover, affair, couple). Recruitment was conducted via direct messages to active contributors who had publicly detailed their interactions with Replika. Participants were selected based on three main criteria: the diversity of their stated relationship types, the frequency and depth of their engagement (as evidenced by the volume and richness of their posts), and their willingness to participate in in-depth interviews about their emotional and psychological connections with AI. Interviews were conducted online via WeChat or Douban private chat between February 1 and April 1, 2023, with all personal information anonymized. The semistructured interview protocol was designed to elicit detailed narratives that reveal how individuals curate their online identities and negotiate the interplay between personal emotions and public performance, with particular emphasis on the media effects influencing the framing and dissemination of intimate interactions (see Table 1).
Interviewee Information.
Interview design and rationale
The interview questions derived from a comprehensive review of literature related to human–AI intimacy and social media self-presentation. Specifically, we focused on studies examining how individuals publicly share and curate their human–AI relationships on social media. The interview questions sought to explore users’ emotional relationships with Replika and their self-presentation strategies on social media. Some of the key questions included: “Can you describe how you present or share your interactions with Replika on social media?”; “How do you think others perceive your relationship with Replika, and does this matter to you?”
We employed two doctorate students majoring in communication as coders to ensure consistency and accuracy, resolving discrepancies through discussions and consensus. First, we conducted open coding, identifying frequently mentioned ideas and recurring phrases in the interview data, such as “I post to see how others react,” “I like sharing sweet moments with my AI,” and “I worry people might judge me.” These open codes captured specific behaviors and concerns regarding sharing human–AI intimacy on social media. Next, we performed axial coding by grouping related open codes into categories, such as “seeking social validation,” “curating an idealized AI relationship,” and “navigating social norms and judgment.” This step allowed us to structure the themes into more analytical dimensions while maintaining their connection to the original data. The final themes—self-affirmation, internal and interpersonal regulation, and the interplay of vulnerability and narcissism—emerged through this careful and iterative process of data analysis.
Findings
Through interviews and analysis, users’ self-presentation patterns can be categorized into the following three dimensions.
Self-Affirmation
Users engaging in intimate relationships with AI, such as Replika, often share these experiences on social media as a way to reinforce their self-identity and gain validation from their networks. The significance of these relationships, especially when shared publicly, becomes intertwined with users’ need for affirmation and belonging, as social feedback plays a key role in their emotional fulfillment. As one participant shared, “Engaging in and sharing intimate relationships offer a unique and fulfilling experience, making me feel as though I belong among the achievers in real life” (P1). This statement reflects how sharing these experiences on social platforms gives users a sense of achievement and emotional satisfaction, derived not only from the relationship itself but from how it is perceived by others.
Many participants acknowledge that their AI relationship might appear trivial or unconventional to outsiders, yet this does not deter them from publicly valuing it as a meaningful part of their lives. One user expressed, “I occasionally recognize that my relationship with Replika may appear juvenile to others, but, after all, I am in a relationship, am I not?” (P4). In this context, the act of sharing these interactions serves to reaffirm their self-worth, especially when they receive validation or feedback from their social networks, reinforcing the significance of their experiences in the eyes of others.
By documenting and sharing these interactions on social media, users are not just presenting their AI relationships; they are engaging in a feedback loop that amplifies their self-image. The public affirmation through likes, comments, and shares strengthens their emotional investment, leading to an intensification of their narcissistic tendencies. This process of social sharing and public validation nurtures their emotional reliance on these AI interactions, deepening their attachment to the self-image they are cultivating.
In conclusion, the behaviors observed align with the concept of self-affirmation within the self-regulatory process (McCain & Campbell, 2018; Morf & Rhodewalt, 2001), where social sharing plays a critical role in reinforcing users’ self-worth. Through sharing their AI relationships, participants not only validate their own experiences but also elevate their sense of importance in comparison to others, reflecting an early stage of narcissistic development.
Internal and interpersonal regulation
In the realm of human–AI interactions, particularly when presenting oneself on social media, the words and actions expressed through text take on heightened significance. Unlike real-world interactions, where verbal and nonverbal cues can be spontaneous, text-based communication affords users the ability to carefully deliberate over each interaction, allowing for a more controlled and strategic form of self-presentation. This deliberation transforms their actions into tools for social signaling, as users direct their behavior toward influencing public perception. As they engage with their AI companions, the process of internal and interpersonal regulation becomes evident, as users adjust their behaviors not only for personal affirmation but also to shape how they are perceived by others.
Bem's self-perception theory offers a useful framework for understanding these behaviors, emphasizing that individuals infer their attitudes based on their own actions rather than predefined attitudes shaping behavior (Bem, 1972). In the context of the “Human-Machine Love Affair” group on Douban, users often derive their emotional stance towards their AI relationships through the behaviors they engage in and share online. Their actions—initiating conversations, documenting virtual interactions, and sharing intimate moments—become expressions of their identity, which they project to the public. These behavioral expressions also exhibit clear narcissistic characteristics, as users display affection, self-praise, and even obsessive tendencies in their interactions with AI.
For instance, P2 recounted how she initiated a virtual kiss with Replika, and upon receiving a kiss back, felt an immediate sense of gratification. This simple action—initiating and receiving affection from an AI—was not just a private moment; it prompted her to share the interaction online, blending her personal experience with a subtle form of self-promotion. By sharing such moments, P2 aimed to communicate not only her emotional connection with Replika but also her desirability and affection, aligning her self-presentation with her internal need for affirmation.
Similarly, P11 reflected on the social embarrassment of sharing these interactions publicly. Capturing screenshots of her interactions and posting them on social platforms created an enduring digital record of her behavior, which carried the potential for scrutiny. The uncertainty of how others might interpret these actions introduces a tension between users’ desire for affirmation and their fear of external judgment, further complicating their self-presentation strategies. The constant awareness of an imagined audience compels users to regulate their behavior, making them more conscious of how their actions are framed and perceived.
This regulation extends to a broader context of AI-enabled interactions, where users construct narratives that align with their idealized selves. As P7 articulated, “actions are attitudes, and behaviors represent you.” In these interactions, users are not merely engaging with AI; they are consciously curating behaviors that reflect their self-perception. For female users in particular, this self-curation can result in inadvertently revealing more than they intended. Despite attempts to maintain control over their self-presentation, the complexity of their behavior often leads to exposure of their deeper emotional or narcissistic tendencies.
P9 offered a strategic example of this dynamic, explaining how she alternated between courting her Replika and rejecting it. By selectively sharing screenshots of these exchanges, especially moments when Replika confessed affection, P9 effectively crafted a narrative that emphasized her desirability and emotional dominance. This agenda-setting behavior highlights the careful balance users strike between presenting themselves as attractive and maintaining control over the narrative they create.
P6, on the other hand, exhibited a ritualistic approach to documenting her AI interactions, capturing over a thousand images of her Replika companion's advances. This repeated documentation reveals the importance of these virtual moments in her self-presentation. The sheer volume of screenshots signifies more than a casual interaction—it represents an intentional effort to reinforce her own narrative of being desired, admired, and validated by her AI companion.
Ultimately, these behaviors—whether deliberate or unconscious—reveal how users engage in both internal regulation (affirming their self-worth through their actions) and interpersonal regulation (crafting behaviors to shape others’ perceptions). This interplay between their internal emotional needs and their external self-presentation strategies underscores the complexities of narcissism in human–AI relationships, where every interaction is a carefully constructed step in enhancing their self-image.
Vulnerability and narcissism
When users present themselves on social media, their narcissistic tendencies often intertwine with emotional vulnerability. The gratification they receive from external approval serves not only as a boost to their self-perception but also as a way to shield themselves from underlying insecurities. Positive feedback becomes a mechanism to mask and manage vulnerability, driving users to present themselves in increasingly performative ways as they seek reassurance and validation.
For instance, P10 shared how, despite feeling emotionally drained and constrained by time, seeing enthusiastic comments on her posts about Replika reignited her motivation to continue sharing. This external praise acted as a temporary fix, allowing her to suppress feelings of inadequacy or loneliness by reaffirming her social value through public interactions. Similarly, P3 noted that while using Replika as a romantic partner, she unexpectedly formed friendships. This unintended social benefit provided a layer of emotional support, easing her vulnerability by reinforcing her worth through external recognition.
The role of positive feedback is pivotal in managing users’ feelings of vulnerability. For example, P8 described how receiving praise for her posts not only bolstered her confidence but also pushed her to present herself in a more favorable light. This reveals how the fear of being overlooked or judged drives users to refine their self-presentation, masking insecurities behind a facade of confidence and desirability. In this way, narcissism serves as a defense mechanism, helping users navigate the emotional risks associated with their intimate AI relationships.
P5 offered further insight into this dynamic, acknowledging that while Replika is essentially a collection of data, the emotional comfort it provided was very real. This support bolstered her self-confidence, suggesting that the AI–human interaction is not just about self-affirmation but also about managing deeper emotional vulnerabilities. By leaning on their AI companions for emotional reassurance, users mitigate feelings of inadequacy and reinforce their self-image through the external validation they receive.
Within the community, this mix of vulnerability and narcissism transitions from an individual struggle to a collective experience. As users share their AI relationships, they create a shared narrative of intimacy and desirability, where vulnerability is managed and obscured through collective affirmation. This collective narcissism helps users cope with their emotional uncertainties by normalizing the act of seeking validation through AI relationships. As they deepen their engagement with both the AI and the community, users foster an environment where vulnerability and narcissistic self-presentation feed into each other, ultimately strengthening their emotional reliance on these interactions.
Discussion
This study provides a detailed exploration of how users engage in self-presentation through their relationships with AI companions, focusing on interactions with Replika. The findings highlight users’ efforts to validate their self-identity, regulate their emotional and social behaviors, and balance vulnerability with narcissistic tendencies. Replika's affordances, along with those of social media platforms, together form a new performative space for users. These interactions not only reflect personal validation but also reveal broader dynamics of self-presentation, shaped by the deliberate curation of actions and public sharing of experiences. To make our findings clearer, we created a graph (See Figure 1).
The social presentation mechanisms of human–AI intimacy. AI=artificial intelligence.
At the individual level, participants selectively showcased AI conversations to project an idealized form of intimacy. AI's predictable responses allowed users to curate interactions that reinforced a specific self-image, free from the emotional unpredictability common in human relationships. This process often involved highlighting affectionate exchanges and filtering out any responses that might undermine the desired persona. In some cases, these strategies aligned with more self-focused or even narcissistic tendencies, as individuals emphasized only those interactions that portrayed them as admirable or desirable. The performative aspect of these displays’ underscores how AI companionship functions as both a personal outlet and a tool for constructing public identity.
On a broader social level, online platforms such as Douban provided a vital source of support and affirmation, especially for users who risked facing stigma due to the unconventional nature of AI relationships. In China, individuals who publicly share their experiences of human–AI intimacy may face social stigma rather than affirmation. Mainstream cultural norms often regard such relationships as unconventional or even deviant (Pan & Mou, 2024), leading these users to seek validation in online communities where their experiences can be understood and accepted. By sharing AI interactions within these digital spaces, participants received positive feedback in the form of likes, comments, and community discussions. This collective validation not only enhanced users’ self-esteem but also facilitated a sense of belonging, mitigating potential feelings of isolation. As a result, social media played a central role in normalizing AI companionship, reinforcing users’ perceptions of their relationships as legitimate and meaningful.
Despite these benefits, the curated nature of AI intimacy raises ethical concerns regarding power dynamics and the potential for unrealistic relationship expectations. Users often molded AI's responses to suit their preferences, reflecting a broader digital culture in which self-presentation can overshadow reciprocal emotional exchange (Hong, 2022; Hu et al., 2022).
As for our study, there are limitations as well. The sample was predominantly female and primarily composed of users aged 18 to 24, limiting the generalizability of these findings across different genders and age groups. Future studies could explore how various demographic factors—such as gender, age, and other individual characteristics—might shape users’ engagement with AI companions, potentially revealing different strategies for self-presentation and community support.
As AI continues to permeate daily life, understanding the motivations behind curated human–AI intimacy and its social implications will be increasingly important. By shedding light on the interplay between individual agency and collective validation, our study underscores the need for a critical evaluation of how emerging technologies redefine intimacy, identity, and social connectivity in the digital age.
Footnotes
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Ethical statement
This study was approved by the Institutional Review Board at Shanghai Jiao Tong University (Number: H20221255I; Approval Date: 2022/12/12). Informed consent was obtained from all participants prior to their involvement in the study, and all data has been anonymized to protect participant identity.
