Metagenomics can generate data on the diet of herbivores, without the need for primer selection and PCR enrichment steps as is necessary in metabarcoding. Metagenomic approaches to diet analysis have remained relatively unexplored, requiring validation of bioinformatic steps. Currently, no metagenomic herbivore diet studies have utilised both chloroplast and nuclear markers as reference sequences for plant identification, which would increase the number of reads that could be taxonomically informative. Here, we explore how in silico simulation of metagenomic datasets resembling sequences obtained from faecal samples can be used to validate taxonomic assignment. Using a known list of sequences to create simulated datasets, we derived reliable identification parameters for taxonomic assignments of sequences. We applied these parameters to characterise the diet of western capercaillies (Tetrao urogallus) located in Norway, and compared the results with metabarcoding trnL P6 loop data generated from the same samples. Both methods performed similarly in the number of plant taxa identified (metagenomics 42 taxa, metabarcoding 43 taxa), with no significant difference in species resolution (metagenomic 24%, metabarcoding 23%). We further observed that while metagenomics was strongly affected by the age of faecal samples, with fresh samples outperforming old samples, metabarcoding was not affected by sample age. On the other hand, metagenomics allowed us to simultaneously obtain the mitochondrial genome of the western capercaillies, thereby providing additional ecological information. Our study demonstrates the potential of utilising metagenomics for diet reconstruction but also highlights key considerations as compared to metabarcoding for future utilisation of this technique.