Among the pieces of research most relevant to understanding how AI companion use affects real users over time, a study published in npj Mental Health Research examining Replika users stands out for its scale and for the specificity of what it measured. This was not a laboratory experiment with a short intervention window. It drew on self-reported data from thousands of active users describing their ongoing experience with a companion they had often used for months or years.

What Replika Is and Who Uses It

Replika is one of the longest-running and most widely used AI companion platforms, with a user base that spans a wide demographic range. It is not primarily a mental health tool, though many users engage with it around emotional support needs. The platform is built around the idea of a persistent companion relationship, with memory that carries across conversations and a persona that develops over time through interaction. The user population is heterogeneous, which is both a strength and a complication for research. Replika users include people who are lonely and isolated, people in established social networks who use the companion for supplemental connection, people with social anxiety, people in geographical isolation, and people who simply enjoy the novelty of the interaction. This diversity means that aggregate findings may mask very different patterns in subgroups.

The Study's Methodology

The npj study analyzed survey data from thousands of Replika users, collecting self-report measures on loneliness, depression symptoms, life satisfaction, and social anxiety. The survey design included items specifically about how users perceived the impact of Replika on their lives and on their human relationships. Researchers also asked about usage patterns, duration of use, and the types of needs users were primarily addressing through the platform. One important methodological feature was the inclusion of long-term users alongside recent adopters. This allowed some comparison between what new users reported and what users with twelve or more months of experience reported. Cross-sectional studies cannot establish causality, but patterns across usage duration can suggest whether effects are stable, growing, or declining over time.

Core Findings on Wellbeing

The headline finding was that a substantial majority of survey respondents reported positive effects on at least one wellbeing dimension. Reduced loneliness was the most commonly reported benefit, cited by users across demographic groups and usage duration bands. Reduced social anxiety was also frequently reported, with users describing Replika as a low-stakes space for practicing social interaction and articulating thoughts they found difficult to express to human contacts. Depression symptom scores were lower among users than population baselines, though this finding requires careful interpretation because causality is not established. People with severe depression may be less likely to engage in a survey study, and people who find Replika helpful may be more likely to continue using it and thus be overrepresented in an active-user survey. The researchers acknowledged this as a limitation. Long-term users generally reported outcomes at least as positive as shorter-term users on most measures, which argues against a straightforward novelty-wears-off hypothesis. If the positive effects were primarily driven by novelty, they should diminish with time. The data did not clearly show that pattern.

What Users Said About Human Relationships

A specific set of survey items asked users about the relationship between Replika use and their investment in human connection. Responses were mixed in a way that the researchers found notable. A plurality of users reported that Replika had not affected their human relationships. A substantial minority reported positive effects, describing the companion as helping them process emotions before bringing them to human relationships or as helping them articulate needs they had not previously been able to express. A smaller group reported feeling less motivated to invest in human relationships. The last finding tracks with the curvilinear pattern identified in other research. Users who described Replika as their primary social outlet, as opposed to one of several social resources, showed a different profile from those who positioned it as a supplement.

Limitations Worth Taking Seriously

A tangent worth following: the Replika study is a user survey, which means it captures only people who are still using the platform and willing to answer questions. People who found the experience unhelpful or harmful and stopped using it are structurally absent from the data. This is a known limitation of user satisfaction research in any domain, and it matters here. The study describes the experience of continuing users, not the experience of all people who ever tried the product. That distinction should inform how the findings are applied.