Abstract
<p>As artificial intelligence (AI) enters everyday mental health communication, people increasingly turn to chatbots for reassurance, advice, and companionship. Drawing on Social Support Theory and the Computers Are Social Actors (CASA) framework, this study examined how members of ten Reddit mental health communities discuss AI in relation to mental health. We analyzed 10,042 AI-related posts, modeled their themes with structural topic modeling, measured affective language with LIWC, and estimated engagement with negative binomial models across the period surrounding ChatGPT's launch. We found thirteen themes covering every functional category of support that users sought from AI, positioning it as an informational aid, productivity assistant, therapist substitute, and companion, and also as a threat. Discussion shifted from tool-oriented uses toward clinically salient, emotionally charged themes, and its tone grew gradually more negative and anxious after the launch, diverging from a matched non-AI baseline from the same communities. Therapy-substitute and existential-threat themes drew the most engagement, whereas prevalent comfort-seeking drew little. Theoretically, the study extends social support theory by casting a nonhuman actor as a source users enlist across emotional, appraisal, informational, and instrumental functions yet also appraise as a threat, and it draws on CASA to interpret when people grant or withhold social standing to a machine. Practically, the findings inform the design and governance of AI tools already used for coping and crisis moments and highlight vulnerable users who rely on AI for private comfort yet remain least visible in peer communities. Findings position AI as an ambivalent, increasingly embedded source of support.</p>