Abstract Rates of mental and behavioral health disorders are increasing among adolescents coinciding with in- creases in overall time spent engaging with social media and other online activities. This trend has led to a strong media narrative that online engagement is spilling over into offline mental and behavioral health. However, empirical evidence is relatively weak and mixed between positive, negative, and null effects. Some evidence suggests there may be benefits from online engagement. For instance, youth who iden- tify as sexual and/or gender minorities (SGM) may benefit from receiving emotional support in desig- nated online spaces that support SGM identities. Other research finds negative impacts for girls’ mental health related to online engagement as well as increased exposure to cyberbullying among SGM youth. There is a strong potential of high impact interventions that consider adolescents embedded within the online context but this area of research is limited by cross-sectional, self-report methods and an almost exclusive focus on time spent online and unidirectional associations rather than bidirectional within-per- son mechanistic processes and rich assessments of online engagement. Passive assessment of online engagement and mental health indicators can provide ecologically valid data on naturalistic adolescent online behaviors and experiences, recognizing that adolescents both shape and are shaped by their online environments. Informed by the Compensatory Internet Use Theory (CIUT), this study leverages multiple data streams (e.g., survey data, ecological momentary assessment (EMA), passive digital trace data from smartphones) to test mechanistic within-person processes that underlie bidirectional links be- tween offline mental and behavioral health and online behaviors and experiences, identifying potentially modifiable targets for future prevention and intervention development. We will recruit both a clinical and non-clinical sample of adolescents age 13-15 (n=450) with an overrepresentation of SGM youth to partic- ipate in the following research activities: (a) three bursts of a 21-day EMA of moods and motivations for engaging online; (b) three pre-burst and three post-burst surveys of mental and behavioral health and online experiences; and (c) passive assessment of digital trace data related to online behavior (e.g., so- cial networking, messaging, music, typed language) and proximal indicators of mental health (e.g., sleep, activity). To address study aims, natural language processing will be utilized to detect hidden mental states in language data and dynamic structural equations modeling will be used to estimate between- person differences in within-person adaptive and maladaptive mechanistic pathways leveraging all data streams. These results are a necessary next step in understanding how real world, ecologically valid as- sessments of online engagement longitudinally and acutely impact mental health and wellbeing as w...