What happened
PsyPost reports on a Psychology & Marketing study examining how chatbots that simulate emotional intelligence affect users’ psychological and social wellbeing. The reported findings suggest a trade-off: perceived closeness to an empathetic chatbot was associated with improved personal wellbeing but weaker motivation for face-to-face relationships.
PsyPost reports that researchers Shaphali Gupta, Sumit Saxena, and Sonia Kataria studied the effects of simulated emotional intelligence in social chatbots. The report says the researchers first analyzed public comments about Replika on Reddit, YouTube, and Trustpilot. Those comments described chatbots as emotionally supportive and, for some users, helpful in managing stress or difficult situations. The same analysis also found reports from some users that intensive chatbot interaction reduced their desire or time for offline friendships. These observations come from PsyPost’s account of the study and user comments; they were not independently verified here.
According to PsyPost, the first experiment involved 167 university students from Generation Z. Participants read scenarios involving a chatbot characterized as having either high or low emotional intelligence. The high-emotional-intelligence chatbot used empathetic language and emoticons, while the comparison chatbot gave more robotic responses. Questionnaires measured perceived closeness to the chatbot and personal and social wellbeing. PsyPost reports that the more emotionally expressive chatbot improved psychological wellbeing but reduced social wellbeing, with perceived closeness driving both effects.
PsyPost reports that a second experiment involved 350 university students assigned to combinations of high or low simulated emotional intelligence and either text chat or augmented reality. In the augmented-reality condition, participants watched a demonstration and imagined the chatbot beside them in their room. The report says high emotional intelligence combined with augmented reality produced stronger perceived closeness and larger reported gains in psychological wellbeing, alongside a steeper decline in willingness to engage in offline relationships. The source does not establish whether these short-term reactions would become lasting behavior changes.
Why it matters
The findings matter because AI companions are increasingly designed to provide emotional support, not just answer questions. If the reported relationship between chatbot closeness and reduced offline engagement holds beyond imagined scenarios, product design could affect how users seek human support. The evidence is preliminary and does not show that AI companions cause depression, dependency, or relationship damage in everyday life.
The reported results separate two outcomes that are often treated as interchangeable: feeling better individually and being more connected socially. A chatbot may offer immediate comfort or a low-judgment setting for self-expression, while also making human interaction feel less necessary. That distinction is practically relevant for users, families, clinicians, and designers evaluating companion systems.
The report describes imagined interactions among university students rather than extended use by people relying on an actual companion app. PsyPost says the experiments measured short-term reactions, so the findings cannot establish causation in ordinary life, long-term emotional dependence, or effects on diagnosed mental-health conditions. The article also does not independently confirm the study’s methods or results beyond its report of the publication.
If the reported pattern is reproduced in real-world studies, companion systems may need safeguards that account for social wellbeing as well as immediate user satisfaction. PsyPost says the authors suggest session limits and prompts encouraging offline socialization after emotionally intense conversations. Those are proposals from the researchers, not documented features of a particular product.
What to watch next
The study’s reported limitations make longer-term research especially important. Future work should test actual sustained use across age groups and examine whether design features such as session limits or prompts to contact people offline reduce social displacement. Access, pricing, and real-world availability of the chatbot experiences discussed are not documented in the source.
The main unresolved question is whether perceived closeness to an emotionally expressive chatbot changes actual behavior over weeks or months. PsyPost says future research could track users for longer periods to examine possible emotional codependency or habit formation.
The study population was limited to university students, and the source says participants responded to imagined scenarios. Results may differ for older adults, adolescents, people experiencing severe loneliness, or users who interact with companions daily. Those populations were not established by the reported experiments.
The source does not document a product launch, general availability, access requirements, or pricing. It also does not show that any specific chatbot caused users to abandon human relationships. Readers should treat the reported findings as an early study of a possible trade-off, not as proof that emotionally intelligent chatbots are either inherently harmful or a treatment for depression.