Background Adherence to medications for treating alcohol use disorder (AUD) is poor. To identify predictors of daily naltrexone adherence over time, a secondary data analysis was conducted of a trial evaluating a mobile health intervention to improve adherence. Methods Participants seeking treatment for AUD (n = 58; M age = 38 years; 71% male) were prescribed naltrexone for 8 weeks. Adherence was tracked using the Medication Event Monitoring System (MEMS). In response to daily text messages, participants reported the previous day's alcohol use, craving, and naltrexone side effects. Using multilevel structural equation modeling (MSEM), we examined baseline dispositional factors and within-person, time-varying factors as predictors of daily adherence. Results Naltrexone adherence decreased over time. Adherence was higher on days when individuals completed daily mobile assessments relative to days when they did not (odds ratio [OR] = 2.53, 95% confidence interval [CI] 1.61 to 3.98), irrespective of intervention condition. Days when individuals drank more than their typical amount were related to lower next-day adherence (OR = 0.93, 95% CI 0.88 to 0.99). A similar pattern was supported for craving (OR = 0.88, 95% CI 0.79 to 0.98). Weekend days were associated with lower adherence than weekdays (OR = 0.71, 95% CI 0.58 to 0.86); this effect was partly mediated by heavier daily drinking (indirect effect = −0.02, 95% CI −0.04 to −0.003) and stronger-than-usual craving (indirect effect = −0.01, 95% CI −0.02 to 0.00) on weekend days.Conclusions The results further demonstrate the need to improve adherence to AUD pharmacotherapy. The present findings also support developing interventions that target daily-level risk factors for nonadherence. Mobile health interventions may be one means of developing tailored and adaptive adherence interventions.