The pervasive integration of AI-driven algorithmic systems into social media platforms has fundamentally altered the digital landscape for adolescents. This review synthesizes longitudinal and multi-country research on the relationship between algorithmic exposure, characterized by personalization, recommendation intensity, and amplification dynamics, and adolescent mental health trajectories. Findings indicate that algorithmically curated content significantly influences psychological well-being through mechanisms such as social comparison, fear of missing out, and sleep disruption. However, these effects are not uniform; they are moderated by cultural context, digital literacy, and socioeconomic factors. Current research remains limited by methodological fragmentation, overreliance on screen-time metrics, and a geographical bias toward high-income countries. This paper proposes a conceptual framework integrating algorithmic inputs, usage patterns, psychological mediators, and contextual moderators to guide future research. The conclusions emphasize the urgent need for longitudinal, interdisciplinary studies across diverse cultural contexts, greater transparency from technology platforms, and policy-relevant investigations to effectively mitigate risks and promote resilience. A global, evidence-based approach is essential to safeguard adolescent mental health in an increasingly algorithmic digital environment.