Restoration of fire-prone forests is a common practice intended to increase resilience to wildfire, drought, and bark beetles. However, the long-term effects of restoration treatments on understory species, particularly non-native species, are poorly understood. We investigated long-term (23 years) effects of restoration treatments on native and non-native understory vegetation at the Lick Creek Demonstration-Research Forest, a ponderosa pine/Douglas-fir forest in western Montana, USA. Restoration treatments included shelterwood-with-reserves harvesting implemented alone or in combination with prescribed burning under either wet or dry conditions, plus a no-action control. Changes in understory vegetation cover and species richness (measured pre-treatment, 2, 3, 4, 5, 15, and 23 years post-treatment) were analyzed by growth form (forb, graminoid, and shrub) and origin (native vs. non-native) to test how understory dynamics respond to disturbance intensity. We also accounted for basal area of trees to consider how differential overstory conditions affected final treatment responses. Native vegetation cover responses followed predictable successional patterns: after initial declines by grasses and shrubs immediately after disturbance (43% and 40% reduction, respectively), all groups increased, peaked about five years after disturbance (74% greater than pre-treatment on average) and then generally declined to levels similar to pre-treatment. The magnitude of these changes was related to the disturbance intensity (i.e., cut-and-dry-burn > cut-and-wet-burn > cut-and-no-burn > control). For all native groups, variation in tree basal area at year 23 explained remaining differences among treatments. Non-native grasses and forbs followed similar overall trajectories, but cover of these groups increased immediately after treatment and rose more steeply to their peak five years post-treatment (12 times greater than pre-treatment on average), with differences persisting through year 23. While non-native cover responses in year 23 were negatively correlated with tree basal area, as seen for native taxa, overstory conditions alone did not fully account for differences between burned and unburned treatments. This suggests that effects of burning unrelated to tree regeneration favored non-natives. Nonetheless, in the last sampling year treatment-induced increases in non-native cover were modest, suggesting that after 23 years understory responses in this forest type were fairly resilient to restoration treatments. However, because forest structure and composition were reverting to pre-treatment conditions by year 23, retreatment may be necessary at intervals of <23-years in our system, which could exacerbate non-native responses. Results of our long-term experiment provide insight into how different forest restoration treatments and their timing may affect understory susceptibility to non-native plant invasion.