Current approaches to modern metabolic profiling of biotechnological samples are facing several challenges, e.g. the wide concentration ranges and the complexity of the matrix. While the information on concentration levels of extracellular metabolites is much in demand, especially for flux balance analysis, the measurement is hampered by a highly complex matrix. This affects the general performance of an analytical method over time. In the present study, an automated clean-up procedure, aiming at matrix separation for the analysis of primary metabolites via liquid chromatography coupled with time-of-flight mass spectrometry, was tested and optimized for supernatants of Chinese hamster ovary cultures. A uniformly 13C labeled yeast cell extract was employed for assessing recovery. The optimized on-line clean-up, comprising filtration and solid phase extraction for matrix removal, showed recoveries ranging from 45.0% to 138.7%, with an average of 11.6% relative standard deviation. Additionally, a non-targeted analytical approach by accurate mass spectrometry for the differential assessment of the matrix separation was performed.