A group of workers internally contaminated with Am have been followed for about 12 years. The source of contamination was AmO2 powder used for production of AmBe neutron sources and other applications. The production of some radionuclide sources included chemical treatment of the original material, which transformed the americium into the nitrate, but mostly powder metallurgy was used for production of sources for smoke detectors. In vivo measurement of the workers was performed with two LEGe detectors placed near the head of the measured person. Calibration was performed with four different physical skull phantoms of different origin and a voxel phantom with Monte Carlo simulation, which was developed to fit the head sizes of individual persons. Samples of urine and feces were analyzed by means of radiochemical separation followed by alpha-spectrometry. Separation of 241Am from mineralized excreta was performed by combined anion exchange and extraction chromatographic techniques. As a tracer, 243Am was used. When the measured data (83 data on skeletal activity, activity in 389 bioassay samples) were compared with International Commission on Radiological Protection's and Leggett's biokinetic models of americium, it was found that in most cases, after more than 15 y since the intake, the excretion rate was lower (or skeletal activity higher) than predicted. On the other hand, the ratio of excreted activity in urine and feces agrees well with model predictions.