Instantaneous intake is central to the understanding of large herbivore foraging strategies and rangeland ecology. Unfortunately, its measurement under field conditions remains challenging because of the difficulty of estimating bite mass. The hand plucking method provides a simple, noninvasive method of estimating bite masses and thus instantaneous intake of grazing herbivores. However, many authors questioned its accuracy and interobserver repeatability. In this study, we tested the accuracy and the repeatability of the hand plucking method using four observers and two herbivore species (i.e., cattle and goats). We compared hand plucked bite mass estimates to actual bite mass of bites taken by the herbivores on natural patches of grass. Training of the observers was fundamental to obtaining accurate bite mass measurements. The mean daily accuracy of the observers' bite mass estimates increased from 60–80% to 80–94% within 5 d. After training, the relationship between bite mass estimates and actual bite mass was linear and not significantly different from a Y  =  X relationship. This means that individual bite mass estimates were centered on the real values and thus positive and negative errors canceled each other when combined. As a result, estimates of cumulative intake over about 10 feeding stations had accuracies of over 95%. Furthermore, neither the observer identity nor the herbivore species affected the accuracy of the measurements. The categorization of bites into different size categories proved to be essential in achieving accurate measurements. When observers are trained, hand plucking is a reliable and accurate method of estimating bites mass and instantaneous intake of grazing herbivores. This has important implications for rangeland research and management, as hand plucking is often the only practicable method available for estimating instantaneous intake of free-ranging herbivores.