In recent years, artificial intelligence (AI) has become deeply embedded in everyday life, managing workplace tasks, scheduling medical appointments, and handling customer support interactions. While these functional uses of AI have been widely accepted, its growing role in emotional and social support-particularly in mental health-has sparked significant debate.

In earlier eras, personal relationships were formed primarily through family, community, and face-to-face interaction; however, a growing number of individuals are now turning to AI companionship, emotional support, and personal connection. This shift occurs within a broader context of loneliness, changing patterns of social connectedness, and increasing comfort with digital platforms. AI-based conversational systems offer users continuous availability, responsive interaction, and opportunities to disclose thoughts and emotions in a private setting. Emerging research suggests that AI companionship may represent a new form of digitally mediated social interaction, particularly among individuals experiencing loneliness or seeking accessible emotional support ^1,2.

From a psychological perspective, AI can provide immediate responses without the fear of embarrassment or rejection and can simulate human-like interaction through language-based models. Research on conversational agents suggests that these systems may reduce psychological distress and provide accessible support for some users ². However, while such responsiveness may offer short-term emotional relief, concerns remain regarding safety, overreliance, problematic use, and the possibility that AI-mediated interaction could alter users’ engagement with human relationships ^1,3. This tension highlights the importance of critically examining not only how AI is used, but also the psychological and social conditions that make such systems emotionally appealing.

At the same time, access to mental health professionals in the United States remains uneven and often insufficient, particularly for vulnerable and underserved populations. Multiple structural and social barriers can limit access to care, including financial constraints, inadequate insurance coverage, workforce shortages, geographic disparities, limited availability of culturally responsive providers, stigma, and other social determinants of health ^4,5. These barriers were further exacerbated by the COVID-19 pandemic, which substantially affected mental health and disrupted the delivery of mental health services ^6,7. In 2022, approximately 50.6% of U.S. adults with any mental illness received mental health treatment during the preceding year, indicating that nearly half did not receive treatment ⁸.

In response to these gaps, AI-based chatbots and conversational applications have emerged as scalable tools designed to provide cognitive behavioral therapy (CBT)-informed interventions and therapeutic-like support. These systems aim to expand access by providing low-cost and readily available psychological assistance, particularly for individuals who may encounter barriers to traditional care. Systematic reviews and meta-analyses indicate that conversational agents can improve certain mental health outcomes, although the magnitude and durability of these effects vary across interventions and populations ^2,9.

Despite their promise, significant limitations remain. Evidence concerning the safety, explainability, cultural appropriateness, and clinical effectiveness of conversational AI remains incomplete, particularly when systems are used for complex mental health conditions or crisis situations ^9,10. In addition, AI systems may reproduce biases present in training data and may provide inappropriate or unsafe responses. These concerns raise important questions regarding oversight, equity, privacy, accountability, and the appropriate scope of AI in mental health support ^10,11.

Given these developments, the objective of this study was to explore the intersection of mental health and AI-based companionship, with a specific focus on how and why individuals use AI systems as a source of emotional support and how these interactions shape perceptions of connection, safety, and well-being.

This study employed a qualitative research design using digital ethnography and thematic analysis to examine the psychological and social factors influencing individuals' reliance on AI for companionship and emotional support. The study sought to explore how users describe their experiences with AI-based companions and therapeutic tools, as well as the underlying motivations that contribute to the development of emotional connections with these technologies.

Data were collected from publicly accessible online discussion platforms between September 2025 and October 2025. The selected platforms included Reddit, Mental Health Forum, and publicly viewable Facebook groups, as these spaces contain naturally occurring discussions where individuals voluntarily share experiences related to AI companionship, chatbot interactions, and AI-assisted emotional support. A purposive sampling strategy was used to identify relevant discussions through keyword searches, including terms such as “AI companion,” “AI friend,” “AI therapist,” “chatbot support,” “emotional attachment to AI,” and “AI for loneliness.” Only English-language posts were included to ensure consistency in data interpretation and analysis.

The study analyzed a total of 150 posts and discussion threads that met the established inclusion criteria. Eligible posts were required to be publicly accessible, written in English, and contain substantive discussion of personal experiences, perceptions, or emotional interactions with AI systems. Posts discussing the use of AI for companionship, emotional support, mental health support, or interpersonal connection were included. Posts were excluded if they focused solely on technical aspects of AI development, software functionality, promotional or marketing content, duplicate discussions, or topics unrelated to emotional or psychological experiences with AI.

Data were analyzed using thematic analysis following the six-phase framework developed by Braun and Clarke ¹². The analysis began with repeated readings of the collected posts to facilitate familiarization with the data. Initial codes were then generated to identify meaningful patterns related to users’ motivations, emotional experiences, perceptions of AI companionship, and the role of AI in addressing loneliness and emotional needs. Codes were reviewed and organized into broader categories, which were subsequently refined into overarching themes through an iterative process.

Attention was given to themes associated with emotional safety, avoidance of interpersonal vulnerability, perceived nonjudgmental support, convenience, accessibility, predictability, and desire for control. Themes were continuously reviewed and compared across the different platforms to identify recurring patterns and variations in user experiences. To enhance the credibility and trustworthiness of the findings, data triangulation was employed by comparing narratives across multiple online communities. An audit trail documenting coding decisions and theme development was maintained throughout the analytical process to support transparency, confirmability, and methodological rigor.

Ethical considerations guided all stages of the research. Only publicly available content was analyzed, and no direct interaction with users occurred. Usernames, profile information, and any potentially identifying details were removed during data extraction and reporting to protect participant anonymity. The study adhered to established ethical principles for internet-mediated research by respecting privacy expectations and minimizing potential risks associated with the use of publicly accessible online data.

Using ethnographic observation on social media, we uncovered two discourses or themes that kept recurring. These discourses were: 1) General usage: Emotional Support & Daily Coping, and 2) Comparisons Between AI-Based Support & Traditional Therapy.

The findings from this study highlight the nuanced ways individuals use AI-based tools such as ChatGPT for emotional support, self-reflection, and daily coping. Users described AI as accessible, immediate, and nonjudgmental, allowing them to process experiences more openly than they might in traditional therapeutic settings. For many, AI served as a supplemental resource that helped them articulate emotions, identify patterns in their thinking, and gain clarity during moments of distress. These findings are consistent with emerging literature suggesting that conversational AI can provide accessible mental health support and may reduce barriers to engaging with digital psychological interventions ^2,9,13,14.

At the same time, a consistent theme across users’ narratives was that AI was not perceived as a complete substitute for human mental healthcare. Rather, participants viewed it as a convenient tool that could enhance-but not replace-the relational, empathetic, and deeply interpersonal aspects of therapy.

Beyond its role as a self-help or therapeutic resource, the findings reveal a more complex phenomenon: the emergence of emotional attachment and companionship between users and AI systems. Many participants described forming meaningful connections with AI, viewing these systems as companions capable of providing comfort, validation, and a sense of connection during periods of loneliness or emotional distress. This trend reflects broader societal changes in which relationships and emotional support are increasingly mediated through digital technologies ^1,3.

Globally, AI companionship is emerging as a significant social phenomenon, particularly among younger populations experiencing loneliness, social isolation, and changing expectations regarding relationships. Recent reporting from China has described the growing popularity of AI-powered companions, including AI boyfriends designed to simulate emotionally intimate relationships ¹⁵. One widely reported example involves Melissa, a Beijing-based human resources manager who described feeling as though she was genuinely in a relationship with an AI companion associated with Xiaoice, a conversational AI platform. Although she recognized that the relationship was not equivalent to a human relationship, she described the companionship as emotionally reassuring ¹⁵.

Such accounts suggest that AI companionship is not merely a technological novelty but may represent a broader cultural response to unmet emotional and social needs. Academic discussions of AI companionship similarly highlight the possibility that AI systems can provide a form of digitally mediated recognition while simultaneously raising questions about whether such interactions fulfill or replace meaningful human relationships ¹.

The growing popularity of AI companions raises important questions about the future of interpersonal relationships and mental health support. While these technologies may provide temporary relief from loneliness, they also offer highly personalized, predictable, and emotionally responsive interactions that differ substantially from human relationships, which require vulnerability, compromise, reciprocity, and negotiation. Existing research on AI companionship and digital loneliness suggests that the social meaning of these interactions cannot be reduced simply to whether they are beneficial or harmful; their effects may depend on users’ characteristics, patterns of use, and existing social relationships ¹.

The findings of this study support this perspective, as participants frequently emphasized the appeal of AI systems that are constantly available, responsive, and free from many of the complexities associated with human interaction. Consequently, AI companionship may be understood not only as a technological development but also as a reflection of broader social trends involving loneliness, changing social connectedness, and dissatisfaction with traditional forms of support ^1,3.

Importantly, participants in this study did not generally view AI as a replacement for professional mental healthcare. Rather, they described AI as a supplemental resource that could assist with emotional processing, journaling, self-reflection, and problem clarification. This distinction is consistent with emerging clinical literature suggesting that AI-based conversational systems may function as adjunctive tools rather than standalone therapeutic interventions ^2,9,14.

While users appreciated the convenience and immediacy of AI interactions, many acknowledged that AI systems lack the empathy, lived experience, clinical judgment, and relational depth associated with human therapists. These findings support the growing view that AI may be most appropriately integrated into existing mental health frameworks rather than positioned as a replacement for professional care ^10,11.

However, these findings must be interpreted within the limitations inherent in analyzing publicly accessible online forums. The study draws from posts written by self-selecting individuals who are comfortable sharing their mental health experiences on social media platforms such as Reddit and Facebook. As such, the perspectives represented may not reflect the broader population of AI users or individuals with mental health conditions. Additionally, posts lack contextual detail, diagnostic verification, and follow-up, making it difficult to fully understand long-term impacts. The anonymity of online spaces also complicates the ability to confirm authenticity, evaluate demographic information, and assess psychosocial factors that shape users’ experiences.

Despite these limitations, online discussions provide meaningful insight into emerging patterns of AI-based emotional support and users’ perceptions of its value.

Beyond methodological limitations, the use of AI for mental health raises serious ethical concerns, particularly around data privacy and regulatory oversight. Many users openly share sensitive psychological information with AI systems without fully understanding how their data may be stored, analyzed, or reused. Importantly, HIPAA protections do not automatically apply to all consumer applications or technology companies. Under U.S. Department of Health and Human Services guidance, HIPAA generally applies to covered entities and business associates, meaning that information entered into consumer applications outside those relationships may not receive HIPAA protection¹⁶.

This creates potential risks related to privacy, third-party access, secondary data use, and data governance.

The literature therefore highlights the need for transparency, accountability, privacy protections, and stronger governance frameworks in digital mental health technologies ^10,11,16.

Notably, the study’s findings align with emerging concerns in academic literature-particularly recent work from Stanford University-which cautions against relying on AI as a standalone therapeutic tool. Stanford researchers reported that AI therapy chatbots may lack the safety, empathy, and situational awareness required for mental healthcare ¹⁷. Their research identified concerning biases toward individuals with conditions such as alcohol dependence and schizophrenia and reported failures in responding appropriately to certain mental health scenarios ^17,18.

These findings underscore the limitations of AI systems in interpreting complex human emotion, assessing clinical risk, and responding with appropriate therapeutic judgment. Conversational AI systems may produce responses that appear empathetic or authoritative while lacking the clinical reasoning and accountability of trained professionals ^10,17.

The Stanford research reinforces a critical distinction articulated by users in this study: while AI may be helpful for reflection, journaling, emotional processing, or gaining quick emotional clarity, it does not possess the relational depth, clinical training, ethical responsibility, or professional accountability required for mental healthcare ^17,18.

AI may nevertheless support mental healthcare through functions such as administrative assistance, appointment scheduling, information navigation, therapist training simulations, documentation, and supplemental emotional support. The appropriate role of AI therefore may be one of augmentation rather than replacement ^11,19.

Taken together, this study suggests that the most effective model for mental health support may be a hybrid approach that integrates AI tools with traditional therapy rather than treating them as competing alternatives. Users valued AI for its immediacy, convenience, accessibility, and nonjudgmental interaction, but also acknowledged that trained professionals provide empathy, accountability, clinical judgment, and human connection necessary for comprehensive treatment.

Pairing AI-assisted reflection with therapeutic guidance may help alleviate some pressures on the mental health system-particularly in settings where access to care is limited-while ensuring that safety, personalization, and ethical oversight remain anchored in human expertise.

Ultimately, while AI offers meaningful benefits as a supportive resource, it should be understood as one component of a broader mental health ecosystem. Its responsible use will require collaboration among developers, clinicians, researchers, policymakers, and users to ensure that technology enhances access to care without replacing the human relationships that remain fundamental to psychological well-being. As AI companionship continues to expand globally, future research should examine the long-term psychological, social, and relational consequences of these interactions to better understand both their benefits and potential risks.

The rise of AI for personal comfort and the development of increasingly intimate relationships with AI reflect a broader shift in how people seek connection, support, and emotional safety in an increasingly digital world. Unlike previous generations, who relied primarily on family, community networks, and in-person interactions, an increasing number of individuals now navigate social isolation, time scarcity, and anxiety surrounding interpersonal conflict through digitally mediated interactions.

AI companions offer an appealing alternative through their ease of availability and emotionally consistent, responsive, and nonjudgmental interactions, while requiring little or no reciprocal effort. For individuals overwhelmed by the complexities of human relationships, AI can provide a controlled and predictable form of interaction that may feel easier and safer than traditional social engagement. However, this convenience may come with an important tradeoff: AI-mediated companionship cannot fully reproduce the reciprocity, vulnerability, empathy, and shared experience inherent in human relationships.

This shift highlights deeper societal issues, including loneliness, changing community ties, and growing comfort with digital intimacy. While AI may support emotional well-being in the short term, excessive reliance on AI could potentially reduce opportunities for human interaction and reinforce social withdrawal. These possibilities warrant further empirical investigation rather than assumptions that AI companionship is inherently beneficial or harmful.

With the continued integration of AI into everyday life, it is crucial to balance its potential benefits with careful attention to its long-term psychological, social, ethical, and relational consequences.