As more people turn to artificial intelligence chatbots for emotional and mental health support, new research from Drexel University suggests that most users view these tools as a complement to human therapy, rather than a replacement. The research, based on millions of posts on Reddit, highlights both the growing appeal of AI support tools for emotional reassurance, coping strategies, and practical guidance, as well as concerns users have expressed about emotional dependence, misinformation, and overreliance on technology.
A 2025 American Psychological Association survey of U.S. adults ages 18 to 80 suggests that nearly half used an AI large language model tool for mental health purposes in the past year. A related study from Brown University found that 1 in 8 young people turn to AI programs for mental health advice. This increased adoption has also raised concerns about the safety and effectiveness of the program.
The Drexel study, led by researchers in the School of Engineering and Computing, sought to learn more about how people use the program and what they think about the responses they get. The findings will be presented at the 2026 Annual Meeting of the Association for Computational Linguistics, an international conference on natural language processing and AI.
Dr Shadi Rezapour, an assistant professor in the university’s School of Computer and Information Sciences, who led the study, said: “Concerns are now growing as people turn to general purpose AI chatbots for mental health support.” “However, these programs are not designed for this purpose and have not been clinically validated. So we wanted to understand how people actually use these systems in their everyday environments and where they see both value and risk.”
Rezapour’s lab studies online narratives, develops socially aware AI systems for marginalized groups, and has been a leader in analyzing Reddit posts to observe the evolution of users’ relationships with various AI programs. For the study, the team used an AI-powered natural language processing program to sift through more than 4 million posts in 47 subreddit groups related to mental health to determine a sample set of 5,126 posts.
By examining the posts using two sociological frameworks (one traditionally used to assess the collaborative bond between therapists and clients, and one used to understand the social acceptance and adoption of new technologies), the researchers were able to answer some important questions about when and how people use AI for therapy and what they think about those interactions.
“We found that people turning to AI for mental health support were often seeking emotional support, empathy, reassurance for anxiety management, coping strategies, and companionship,” Rezapour said. “We also observed that many users relied on these tools for practical support, such as organizing and managing tasks related to attention-deficit/hyperactivity disorder and autism.”
User attitudes toward the program reflect cautious misgivings about this technology, with 51% of posts explicitly mentioning the program’s risks and limitations when used in a therapeutic context. Many people expressed concerns about becoming addicted or psychologically dependent on the program.
And in nearly every use case, our analysis showed that more users frame AI programs as complementary to human therapists, rather than as better or worse than them.
“What we found in our research is that very few people are using AI to replace treatment,” Rezapour says. “More often, they described using it alongside treatment or when human care is unavailable, inaccessible, or inadequate.”
The study also revealed what researchers call the “bond paradox.” Users reported more positive experiences when AI assisted them with specific tasks and goals, such as reflecting, coping, and organizing. However, strong emotional bonds with AI were often associated with risks when not combined with clear goals or useful tasks, especially in the pursuit of companionship and repeated reassurance. In these cases, users often spoke of dependence, worsening of symptoms, feelings of shame and guilt, and increased emotional dependence on the system.
“Our findings suggest that AI tools should not be designed to simply feel warm or human-like,” said Elham Aghhani, lead author of the study and a doctoral student at the university. “Particularly in use cases where companionship or repeated reassurance is sought, clear boundaries and safeguards are needed, as users more frequently describe dependence, worsening of symptoms, and difficulty with withdrawal.”
Taken together, these results suggest that well-designed AI programs for mental health should be optimized for supportive interactions with clear boundaries, rather than emotional bonds or peer relationships, according to the researchers.
“While our findings suggest that many users approach these systems with some degree of caution, it remains important that AI tools designed to support mental health are based on established, evidence-based frameworks,” Rezapour said. “As these technologies become more widely adopted, it becomes increasingly important that users understand both their potential benefits and limitations.”
Read the full study here: https://arxiv.org/abs/2601.20747
