https://doi.org/10.5281/zenodo.18771968

Artificial intelligence (AI)-driven recommendation algorithms have become central to the user experience on major social media and entertainment platforms, including YouTube, Instagram, TikTok, Facebook, and Netflix. By curating personalized content feeds, these systems profoundly shape what billions of people see, think, and feel on a daily basis. This paper presents a comprehensive empirical investigation into the mental health implications of AI recommendation algorithms, examining three primary outcome dimensions: addiction (compulsive, dependency-driven use), anxiety (heightened stress, social comparison, and fear of missing out), and assistance (positive outcomes such as access to educational content, self-help resources, and community support).Using a mixed-methods cross-sectional design, this study collected data from 1,247 participants across diverse age groups (13-65+), professional backgrounds, and geographic regions. Quantitative instruments included the Generalized Anxiety Disorder-7 (GAD-7) scale, the Bergen Social Media Addiction Scale (BSMAS), and a custom Algorithmic Assistance Index (AAI). Qualitative data were gathered through structured interviews with 45 participants representing distinct demographic cohorts.Findings reveal a statistically significant positive correlation between high daily algorithmic exposure and both compulsive usage patterns (r = .61, p < .001) and anxiety symptoms (r = .47, p < .001). Adolescents and young adults showed heightened vulnerability to negative mental health outcomes. Paradoxically, approximately 28% of participants reported meaningful positive mental health effects from algorithmically curated content, particularly in domains of health information access, emotional support communities, and skill development. These findings underscore the dual nature of AI recommendation systems and call for evidence-based policy interventions, platform design reforms, and public digital literacy initiatives.