Loneliness and AI: Do Chatbots Replace Human Relationships or Complement Them?
In recent years, AI-powered tools such as ChatGPT or Gemini have spread rapidly and are now present in many areas of life. Increasingly, such systems are being used for personal conversations, for example about hobbies, travel, everyday problems, mental health, or current events. This development raises a central societal question: do such conversations with AI replace human relationships, or do they provide a meaningful complement? On one hand, there is concern that people may retreat further into digital worlds and neglect real social contacts, which could exacerbate loneliness, social isolation, and withdrawal. On the other hand, there is evidence that AI systems can provide emotional relief and help people better cope with personal or social challenges.
Previous research has focused primarily on social robots or specific groups such as students or older adults. Early recent studies indicate that AI-powered conversational systems can reduce feelings of loneliness and isolation by providing support, stimulation, and a sense of connection. At the same time, comprehensive studies on the general adult population and the targeted use of freely available AI tools for personal conversations are lacking. This research gap is problematic because social connectedness plays a central role in health. Loneliness and social isolation are closely associated with higher morbidity and mortality rates. The German Institute for Economic Research (DIW Berlin) shows that one in five people in Germany sometimes feel lonely. Different facets of loneliness can be distinguished. The most commonly affected are people experiencing solitude (56%), i.e., missing other people, whereas feelings of social isolation (20%) or exclusion (28%) occur less frequently. Various factors such as low income, gender, or age influence the occurrence of loneliness.
Current research aims to examine whether using AI for personal conversations is associated with loneliness, perceived social isolation, and social withdrawal, and whether certain population groups are particularly affected. Fundamentally, two opposing perspectives can be distinguished: on one hand, AI systems could provide emotional support, reduce stress, and strengthen the sense of belonging, similar to human social support. On the other hand, there is concern that AI could partially replace real relationships, thereby promoting social alienation in the long term. Considering that loneliness is a serious global health issue, linked to depression, anxiety disorders, cardiovascular diseases, and increased mortality, it becomes clear how important it is to better understand the influence of AI-based social interactions.
The study “Association of using AI tools for personal conversation with social disconnectedness outcomes” investigated, using a representative online survey of 3,270 participants aged 18 to 74 in Germany, how the frequency of AI tool usage exclusively for personal conversations is associated with social feelings such as loneliness, perceived social isolation, and social withdrawal. Established measurement instruments were used, particularly the 6-item Loneliness Scale by De Jong Gierveld, which ranges from 0 to 6, with higher scores indicating greater loneliness. The average score in the total sample was 3.3 points (standard deviation ± 2.0), reflecting a moderate level of loneliness.
Results by usage frequency showed that around 48% of respondents never used such AI tools for personal conversations, 27% used them one to three times per month or less, and 25% used them at least once per week. Individuals who used the tools one to three times per month or less showed only a slight increase in loneliness compared to non-users (β = +0.14, p = 0.096). The effect was more pronounced among weekly users, who reported, on average, about 0.37 points higher on the 0–6 scale (β = +0.37, p < 0.001). This indicates that frequent use of AI for personal conversations is associated with stronger perceived loneliness. Further analyses showed that gender and age moderated these associations. Negative effects on social isolation and withdrawal were more pronounced in men and decreased with age, while educational level played no significant role. The authors emphasize that these are cross-sectional data, so it cannot be determined whether frequent AI use causes social isolation or whether socially isolated people are more likely to use AI tools.
The study “Artificial intelligence chatbots as a source of virtual social support: Implications for loneliness and anxiety management” followed an experimental approach to investigate how different types of support messages from social AI chatbots influence the perception of emotional validation, especially in the context of loneliness and anxiety. Personalized messages were compared with less personalized ones. Results showed that highly tailored messages led to increased emotional validation. Participants felt more understood and affirmed by these empathically formulated AI responses. This effect was partly mediated by the perceived quality of social support and the subjective sense of interpersonal warmth. Additionally, the perception of social presence of the AI—i.e., how “real” and embedded in social contexts the interaction appeared—played a role. The study demonstrates that AI chatbots can potentially act as a source of social support through targeted, person-centered communication, especially where human support is not available. At the same time, the long-term effects and limitations of such virtual support systems still need further investigation.
The study “A Study of Social Chatbots Affordances Mitigating Loneliness” by Wang and colleagues analyzed how social chatbots can reduce loneliness and which mechanisms are involved. The research model is based on Leonardi’s affordance theory and Tamir’s implicit theories of emotion. Two central affordances of chatbots were considered: Shared Identity, the ability of the chatbot to convey its own social identity, enabling users to feel connected, and Social Support, functions that provide emotional support, recognition, belonging, and advice. Both affordances can directly reduce loneliness and indirectly through the intimacy generated by the chatbot, i.e., trust and willingness to share personal thoughts. Another central factor is the users’ emotion regulation. People who believe emotions are changeable respond more strongly to chatbot affordances, whereas those who view emotions as fixed show smaller effects.
Methodologically, the study used a one-month longitudinal design, in which students were randomly assigned to one of three chatbots: one with Shared Identity affordance, one with Social Support affordance, or one with both affordances. Loneliness, intimacy generated by the chatbot, perception of affordances, and emotion regulation were measured at two time points. The study makes important contributions by systematically showing which chatbot affordances reduce loneliness and foster intimacy and by developing a model explaining the mechanisms of human-chatbot interactions. Practically, it provides insights for designing social chatbots that can support people with mental health challenges, especially in contexts with a shortage of professionals or stigmatization.
Overall, the studies present an ambivalent picture: while AI-powered conversations can provide emotional support and convey social closeness, there is evidence that frequent use is associated with higher loneliness scores and greater social withdrawal. Research suggests that the success of AI interactions strongly depends on system design, personalization, perceived social affordances, and individual psychological factors. A deeper understanding of these relationships is crucial to developing AI systems that meaningfully complement human relationships rather than replace them.
Sources:
Berlin, D. I. W. (2025). DIW Berlin: Einsamkeit in Deutschland : die gefährdetste Gruppe sind Menschen mit niedrigem Einkommen [Text]. DIW Berlin. Source
Hajek, A., Zwar, L., Gyasi, R. M., Yon, D. K., Pengpid, S., Peltzer, K., & König, H.-H. (2025). Association of using AI tools for personal conversation with social disconnectedness outcomes. Journal of Public Health. Source
Merrill Jr., K., Mikkilineni, S. D., & Dehnert, M. (2025). Artificial intelligence chatbots as a source of virtual social support: Implications for loneliness and anxiety management. Annals of the New York Academy of Sciences, 1549(1), 148–159. Source
Wang, W., Sun, H., & Miranda, S. (2024). A Study of Social Chatbots Affordances Mitigating Loneliness. SIGHCI 2023 Proceedings. Source
Yang, Y., Wang, C., Xiang, X., & An, R. (2025). AI Applications to Reduce Loneliness Among Older Adults: A Systematic Review of Effectiveness and Technologies. Healthcare, 13(5). Source