Headspace GC-MS has been opitimized for the determination of furan in foods. The conditions of sample preparation, headspace sampling and GC separation were optimized to enhance sensitivity during GC-MS analysis. Green coffee was used to prepare a matrix matched calibration curve for furan. However, it was unexpectedly found that a green coffee sample was not blank. GC-MS analysis performed after equilibration for 30 min at 40°C showed the presence of 4.2 ng/g furan in green coffee. In order to understand whether furan was naturally present or formed during headspace sampling, green coffee was investigated in time-dependent manner at headspace equilibration temperatures of 40 and 70°C. It was observed that furan response continued to increase in a way similar to first order formation kinetics. The same behavior was found for freshly squeezed tomato and orange juices leading to the suspicion of furan formation during headspace equilibration. It is concluded that a matrix matched calibration for each particular food matrix is necessary to compensate for furan formation during headspace sampling, and thus, to quantify furan more accurately.