Alcohol screening, brief intervention and referral to treatment (SBIRT) in adult primary care is an evidence-based, public health strategy to address unhealthy alcohol use, but evidence of effectiveness of alcohol brief intervention (ABI) in real-world implementation is lacking. We fit marginal structural models with inverse probability weighting to estimate the causal effects of ABI on 12-month drinking outcomes using longitudinal electronic health records data for 312,056 adults with a positive screening result for unhealthy drinking between 2014 and 2017 in a large healthcare system that implemented systematic primary care-based SBIRT. We examined effects of ABI with and without adjusting for receipt of specialty alcohol use disorder (AUD) treatment, and whether effects varied by patient demographic characteristics and alcohol use patterns. Receiving ABI resulted in significantly greater reductions in heavy drinking days (mean difference [95% CI] = -0.26 [-0.45, -0.08]), drinking days per week (-0.04 [-0.07, -0.01]), drinks per drinking day (-0.05 [-0.08, -0.02]) and drinks per week (-0.16 [-0.27, -0.04]). Effects of ABI on 12-month drinking outcomes varied by baseline consumption level, age group and whether patients already had an AUD, with better improvement in those who were drinking at levels exceeding only daily limits, younger, and without an AUD. Systematic ABI in adult primary care has the potential to reduce drinking among people with unhealthy drinking considerably on both an individual and population level. More research is needed to help optimize ABI, in particular tailoring it to diverse sub-populations, and studying its long-term public health impact.