Neural measures of alcohol cue incentive salience have been associated with retrospective reports of riskier alcohol use behaviour and subjective response profiles. This study tested whether the P3 event-related potential (ERP) elicited by alcohol-related cues (ACR-P3) can forecast alcohol use and craving during real-world drinking episodes. Participants (N =262; Mage = 19.53; 56% female) completed a laboratory task in which they viewed images of everyday objects (Neutral), non-alcohol drinks (NonAlc) and alcohol beverages (Alc) while EEG was recorded and then completed a 21-day ecological momentary assessment (EMA) protocol in which they recorded alcohol craving and consumption. Anthropometrics were used to derive estimated blood alcohol concentration (eBAC) throughout drinking episodes. Multilevel modelling indicated positive associations between P3 amplitudes elicited by all stimuli and within-episode alcohol use measures (e.g., eBAC, cumulative drinks). Focal follow-up analyses indicated a positive association between AlcP3 amplitude and eBAC within episodes: Larger AlcP3 was associated with a steeper rise in eBAC. This association was robust to controlling for the association between NonAlcP3 and eBAC. AlcP3 also was positively associated with episode-level measures (e.g., max drinks, max eBAC). There were no associations between any P3 variables and EMA-based craving measures. Thus, individual differences in neural measures of alcohol cue incentive salience appear to predict the speed and intensity of alcohol consumption but not reports of craving during real-world alcohol use episodes.