The aim of this study was to estimate the lowest concentration of pesticide residues in non-fatty food matrix at which the residues can be successfully identified by automatic spectral deconvolution software AMDIS. For GC-MS measurements fast GC with narrow capillary column was utilized. For a mixture of 18 pesticides, the identification was successful at concentration levels 4–0.4 mg kg⁻¹ in real matrix samples (apples). With decreasing concentration, the number of identified pesticides and the quality of deconvoluted spectra decreased. The calculated limits of full-scan detection ranged from 0.20 ng for chlorpyrifos to 1.10 ng for captan. Software AMDIS with the used experimental set-up is not sufficiently sensitive for reliable identification of pesticide residues in non-fatty food matrices with low maximal residual limits (e.g. baby food 0.01 mg kg⁻¹).