The purported sustainability of sustainable-use reserves (SURs) has been questioned in recent decades due to anthropogenic disturbance, including widespread game hunting. A fuller understanding of the drivers of harvest-induced game population changes in SURs is needed to inform this debate. We deployed 720 camera traps around 100 local communities both inside and outside nine SURs in central-western Brazilian Amazonia to generate detection records of 29 mammal and bird species. We used Royle-Nichols multi-species occupancy models to evaluate if (i) distance to and size of local communities, (ii) local human population density, (iii) distance to and size of urban areas, (iv) local level of protection, and (v) alternative availability of aquatic protein affected the (a) species richness, (b) aggregated abundance and (c) biomass, (d) mean reproductive rate of species, and (e) mean abundance of functional groups and (f) individual species. Community distance was the main determinant of wildlife declines, impacting species up to 5-km from communities, but three game species exhibited higher abundances within this distance. Other drivers, such as community size and urban neighbourhood, also contributed to species declines. Availability of alternative aquatic protein buffered declines of only two species and local protection increased species richness and aggregate abundance. These findings can help inform evidence-based conservation strategies in tropical SURs. Our results suggest that preventing habitat loss beyond 5-km radius from communities can promote a healthy source-sink dynamic for populations of game species. Furthermore, game management measures could encourage targeting harvest-tolerant species and the protection of all game species.