The aim of the present study was to compare body fat mass (FM) and fat-free mass (FFM) estimates by bioelectric impedance spectroscopy (BIS), with respective estimates by dual-energy X-ray absorptiometry (DXA), in obese and non-obese subjects. Body composition was measured in ninety-three obese and non-obese men and women by BIS device, BodyScout (Fresenius Kabi, Bad Homburg, Germany) and DXA device, Lunar iDXA (GE Healthcare, Madison, WI, USA). Mean difference between the methods was analysed by t tests, and Bland–Altman plots were generated to further examine the differences between the methods. Mean difference between the estimates by DXA and BIS (ΔDXA − BIS and Bland–Altman 95 % limits of agreement) were as follows: FM 4·1 ( − 2·9, 11·2) kg and 4·5 ( − 2·9, 11·8) %, FFM − 4·1 ( − 11·2, 2·9) kg and − 4·5 ( − 11·9, 2·9) %, indicating large inter-individual variation and statistically significant underestimation of FM and overestimation of FFM by BIS, as compared to DXA. The underestimation of FMkg (FM measured in kg) and overestimation of FFMkg (FFM measured in kg) were more pronounced in men than in women, and the underestimation of FM% (FM measured in percent) and overestimation of FFM% (FFM measured in percent) were more pronounced in normal weight (BMI = 20·0–24·9 kg/m2) than in overweight and obese (BMI ≥ 25·0 kg/m2) subjects. BIS may be suitable for classification of a population into groups according to FM and FFM. However, the large inter-individual variation suggests that this BIS device with the proprietary software is insufficient for estimation of single individual body FM and FFM.