Feed is the main incentive to encourage voluntary cow traffic in automatic milking systems (AMS). Pasture-based AMS farmers utilize access to pasture and concentrate in the robot to manage cow traffic and ensure target milk yield (MY). We hypothesized that variability in concentrate intake (CI) fed in the robot affects negatively MY, which might reduce AMS profitability. A database from 17 pasture-based AMS farms from Australia, New Zealand, and Ireland was used. We analyzed 403,226 daily records over a 5-mo period (September 2018 to January 2019) including cow id, stage of lactation (SOL; early = 0 to 100, mid = 101 to 200, late = >201 d in milk), parity (primiparous vs. multiparous), MY (kg/d), milking frequency (MF, milkings/d), and concentrate intake (CI, kg/d as fed). We calculated average and coefficient of variation (CV) for each variable for a fixed 7-d period. We ran partial correlations (r; calculated using the MANOVA / PRINTE commands of PROC GLM of SAS Version 9.3, SAS Institute Inc., Cary, NC) to evaluate association between MY and CVMY and the other variables (CI, MF, CVCI, CVMF). The model included farm, month, SOL, and parity as fixed factors. As expected, MY was highly and positively correlated (P < 0.01) with CI (r = 0.629) and MF (r = 0.585). Milking frequency was also highly and positively correlated with CI (r = 0.529). However, CV of CI and MF affected negatively both parameters. Milk yield was highly and negatively correlated (P < 0.01) with CICV (r = - 0.249) and MFCV (r = - 0.229). Milking frequency was also highly and negatively correlated (P < 0.01) with CICV (r = - 0.247). Our data show that milk yield in pasture-based AMS is negatively affected by variability in concentrate intake in the robot. Therefore, reducing concentrate intake variation is key to maximize milk yield output.