In order to identify tea varieties quickly, efficiently and nondestructively, an adaptive improved possibilistic c‐means (AIPCM) clustering with the fuzzy Mahalanobis distancewas proposed to classify the Fourier transform infrared reflectance (FTIR) spectra of tea samples.Three varieties of tea samples were scanned andFTIR spectra were acquired for 96 tea samples of different geographical originsby using the FTIR‐7600 infrared spectrometer. Multiple scatter correction (MSC) was employed to eliminate light scattering of FTIR spectral data. After that, principal component analysis (PCA) and linear discriminant analysis (LDA) were applied to reduce the dimensionality of FTIR spectra and extract the discriminant information, respectively. Thenthe data were clustered byseveral fuzzy clusteringalgorithms. AIPCM had the highest clustering accuracyand its accuracy achieved up to 98.5%. The experimental results showed that FTIR spectroscopy coupled with AIPCM clustering was superior in classification of tea varieties.