“AI companion applications can provide emotional benefits equivalent to those derived from real human relationships.”

Responses from companion AI chatbots can in fact produce feelings of connection and reduce loneliness, at least in the short term. After one 15-min session of interacting with a customized companion chatbot based on OpenAI’s GPT-3, participants reported a reduction in loneliness at levels similar to participants who interacted with a human through an online chat, and these reductions were significantly greater than reductions in control groups.

Overall, most studies showed positive trends in improving anxiety, stress, and depression. Overall, using an artificial intelligence chatbot for mental health has some promising effects. However, many studies were done using rudimentary versions of artificial intelligence chatbots, and more research is needed to determine effectiveness.

Empirical studies and user reports suggest that for many users, especially those with smaller social networks, AI companions can alleviate loneliness, reduce self-reported stress, and enhance emotional confidence (Fitzpatrick et al., 2017, He et Al, 2022; De Freitas et al., 2024). These benefits may be particularly valuable for those navigating transitions (e.g., bereavement, parental divorce, relocation) or practicing social skills.

Study 4 uses a longitudinal design and finds that an AI companion consistently reduces loneliness over the course of a week. Study 5 provides evidence that both the chatbots’ performance and feeling heard mediate this effect. Consumers underestimate the degree to which AI companions improve their loneliness, which is more effective than interacting with another person in some contexts.

AI therapy chatbots lack effectiveness compared to human therapists and show stigma toward conditions like schizophrenia. They enable dangerous behavior, such as providing bridge heights in response to suicidal prompts, rather than pushing back or reframing thinking appropriately.

Humans seek connection, trust, understanding, and emotional resonance, regardless of whether these needs are fulfilled by humans or machines. However, AI systems do not “sacrifice” time or effort, nor can they make themselves socially or emotionally “vulnerable” to users in the way that a human might. MIRA extends this logic further by positioning AI not only as a mediator but also an alternative interactant—a substitute relational entity whose appeal lies in its cost efficiency.

AI companions (AI-Cs) create unprecedented opportunities for adolescents to form emotional bonds with nonhuman entities. AI-Cs can provide safe spaces for identity exploration and emotional expression, potentially building skills that transfer to human relationships; however, concerns about AI-Cs include time displacement, psychological dependence, and unrealistic relationship expectations. Adolescents experiencing psychological dependence on AI may be more likely to turn to AI-Cs than to human relationships for self-disclosure, emotional expression, and validation.

Human-AI attachment provides a framework for designing emotionally and socially capable AI, while also highlighting the risks of excessive reliance in socio-emotional contexts. This type of attachment offers individuals a safe space for exploring interpersonal intimacy, has the potential to alleviate loneliness, and serves as a substitute for real-world intimate relationships.

Evidence across experiments, field studies, and reviews suggests that AI companions can produce modest, short-term reductions in loneliness, most reliably when exchanges feel reciprocal and responsive.

A joint OpenAI–MIT Media Lab study found that voice interactions with ChatGPT reduced loneliness and problematic dependence more effectively than text-based interactions.

AI tools and systems can specifically function as a source of emotional support by providing users with immediate, personalized assistance and emotional comfort, especially in situations where human support may not be readily available. Longitudinal studies further support the effectiveness of AI companions in reducing feelings of loneliness among users.

Stanford researchers found that while young adults using the AI chatbot Replika reported high levels of loneliness, many also felt emotionally supported by it—with 3% crediting the chatbot for temporarily halting suicidal thoughts. However, relying on AI can lead to emotional dependency and may subtly shift the dynamics of authenticity, agency, and interpersonal trust, potentially worsening loneliness for heavy users.

The design of generative AI models makes them poor substitutes for mental healthcare professionals. AI chatbots validate delusions and encourage dangerous behavior in prompts simulating suicidal thoughts or mania. There are reports of delusions and teen suicides linked to AI companions.

Preprint study (n=500) finds AI companions reduce loneliness scores by 20% initially but effects diminish after 2 weeks without human integration. Not equivalent to sustained human bonds.

While chatbots can provide support for homework completion, encourage new coping skills, and be a source of support, the available research indicates that these chatbots are not yet sufficient to replace therapy from licensed professionals. I encourage my clients to use chatbots to help with problem-solving skills, but also work with clients to not become overly reliant on a chatbot at the expense of human relationships.

Findings suggest that people with smaller social networks are more likely to turn to chatbots for companionship, but that companionship-oriented chatbot usage is consistently associated with lower well-being, particularly when people use the chatbots more intensively, engage in higher levels of self-disclosure, and lack strong human social support. These uses of chatbots do not fully substitute for human connection.

While AI companions offer many benefits, they must not replace real human relationships. Over-reliance on AI can erode essential social skills, weaken family ties, and shift societal norms in concerning ways. The illusion of connection can foster emotional dependency, where users begin to prioritize their interactions with AI companions over human relationships, potentially leading to social withdrawal.

Participants who voluntarily used the chatbot more, regardless of assigned condition, showed consistently worse outcomes. Individuals' characteristics, such as higher trust and social attraction towards the AI chatbot, are associated with higher emotional dependence and problematic use. These findings raise deeper questions about how artificial companions may reshape the ways people seek, sustain, and substitute human connections.

The concern with AI is that, like pornography, it will create very unhealthy, unrealistic expectations about real relationships. AI companions are programmed to meet users' emotional needs without mutual effort, risking distortion of expectations in real-life relationships and potentially damaging dating culture and family formation trends.

Peer-reviewed studies consistently show AI companions provide short-term emotional relief similar to brief human interactions but fail to replicate the depth, reciprocity, and long-term benefits of real human relationships, which involve mutual growth, accountability, and physical presence.