In this work, surface-enhanced Raman scattering (SERS) spectrum was used to build a rapid analytical method for pesticide residues in fruits. QuEChERS (quick, easy, cheap, effective, rugged and safe) method was modified by multi-walled carbon nanotubes as clean-up sorbents. Overall recoveries of selected pesticide phosmet ranged from 77 to 97% in apples at three spiking levels (0.5, 1 and 2 mg kg −1). The relative standard deviations were between 6.6 and 14%. The limit of detection for phosmet was 0.1 mg kg −1 in standard solution and 0.5 mg kg −1 in apples, which was below the maximum residue limits in fruits of USA, EU and China. The intensity of phosmet characteristic peak existed good linear relationship with the logarithm of concentration between 0.5 and 5 mg kg −1 , with the calibration curve coefficients (R 2) of 0.9994, which indicated quantitative potential for pesticide residue detection. The method was extended to other pesticides, and the obtained SERS results could be used to establish a spectra database. All the experiments were performed with a portable Raman instrument. Combining pretreatment method with spectra database, a sensitive, rapid and convenient method could be built for pesticide residues detection in fruits.