The study of diet is central to wildlife ecology, management and conservation. Metabarcoding increased the capability to identify species contributing to wildlife diet and blocking primers can maximize the detection of prey. Relative read abundance (RRA) of different prey species has been used as semi-quantitative approach, assuming that RRA reflects species contribution to diet. However, this approach has been contested and it is unclear how blocking primers might affect the result. We tested accuracy of RRA to estimate diet by feeding captive wolves six different diets. We analyzed samples without and with four blocking primer concentrations (5, 10, 15 and 20x) to provide insight into the validity of the semi-quantitative metabarcoding approach and the influence of blocking primer in results. RRA produced a highly accurate representation of actual contribution to diet, that was best without blocking primer (0.775 ± 0.033; P <0.001; R2=0.815) with no difference in the number of diet items detected when compared with analyses with blocking primer. Adding blocking primer resulted in higher proportions of reads of diet items, as opposed to wolf sequences, but did not increase the probability of detecting diet components, increased detections of items not fed to wolves, and produced slightly less accurate estimates of diet composition. Finally, resampling suggested that sample sizes beyond 30 scats reduced the variation in results. While our results are promising and support the use of metabarcoding to determine volumetric contribution of items to diet, caution and further research are needed before safe extrapolation to filed studies.