The rising presence of wildfires in western US landscapes necessitates a better understanding of how wildlife respond to these disturbances both immediately and as scorched habitats recover. Managing for species, such as deer, requires an understanding of which forage items sustain them over different post-fire phases. We used molecular genetic analysis of feces (metabarcoding) to document changes in deer diet during springs of 2017–2021, spanning before and after a large wildfire (Ranch Fire of 2018) in Northern California. We used 707 pellet samples collected both on and off the burn area throughout the 5 years, providing pre-fire (2 yrs), post-fire (1 yr), and recovery (2 yrs) periods, each with burn and control comparisons. We predicted that dietary diversity would decrease post-fire in the burn area as herbaceous and shrub layers were reduced, and increase during the recovery period as pioneer species began to recover. We extracted DNA from fecal pellets and sequenced a trnL gene region of the chloroplast genome on a high-throughput platform. We identified 90 plant taxa from an average of ~7,000 sequencing reads per fecal sample. Diet diversity was high in pre-fire and recovery periods in all sites but significantly reduced post-fire in burn sites only (P < 0.001), with diet dominated by a single forage item, oak (Quercus spp). Diet diversity was dominated by shrub taxa pre-fire but composed proportionally more of herbaceous taxa during recovery. Thus, oak species played a central role in buffering the deer population from immediate impacts of wildfire.