Monitoring the industrial production of galacto-oligosaccharides (GOS) requires a fast and accurate methodology able to quantify, in real time, the substrate level and the product yield. In this work, a simple, fast and inexpensive UV spectrophotometric method, together with partial least squares regression (PLS) and artificial neural networks (ANN), was applied to simultaneously estimate the products (GOS) and the substrate (lactose) concentrations in fermentation samples. The selected multiple models were trained and their prediction abilities evaluated by cross-validation and external validation being the results obtained compared with HPLC measurements. ANN models, generated from absorbance spectra data of the fermentation samples, gave, in general, the best performance being able to accurately and precisely predict lactose and total GOS levels, with standard error of prediction lower than 13 g kg(-1) and coefficient of determination for the external validation set of 0.93-0.94, showing residual predictive deviations higher than five, whereas lower precision was obtained with the multiple model generated with PLS. The results obtained show that UV spectrophotometry allowed an accurate and non-destructive determination of sugars in fermentation samples and could be used as a fast alternative method for monitoring GOS production.