Sustainable management of groundwater resources requires a comprehensive understanding of the groundwater system and its water balance, including groundwater extraction. Bores (otherwise known as wells) used for cattle grazing are not typically metered and extraction is usually estimated using analytical models based on supply and demand-based factors. A world-first program of metering of 36 bores on 24 beef cattle grazing properties and landholder interviews on factors affecting water use was undertaken in semi-arid southern Queensland, Australia, over a 6-year period. This study investigated whether data from this program could be used to develop empirical models that could improve estimates of groundwater extraction. A model to predict cattle numbers was initially developed. The outputs from this model, together with public domain spatial data sets, were then used to predict groundwater extraction using a generalized linear model. Modeled property groundwater extraction ranged from 0.31 to 25.5 ML a-1 and was 7.6 ML a-1 on average. The model explained 83.3% of the variance in metered extraction and was found to be more accurate for the metered properties than models previously applied. The study demonstrates that models developed using data for a representative subset of users can be an effective means of estimating groundwater extraction. The approach could be applied across a range of hydrological environments and agricultural settings globally to improve estimation of unmetered take.