- Two solutions, at opposite ends of a continuum, have been proposed to limit negative impacts of human agricultural demand on biodiversity. Under land sharing, farmed landscapes are made as beneficial to wild species as possible, usually at the cost of lower yields. Under land sparing, yields are maximised and land not needed for farming is spared for nature. Multiple empirical studies have concluded that land-sparing strategies would be the least detrimental to wild species, provided the land not needed for agriculture is actually spared for nature. However, the possibility of imperfections in the delivery of land sparing has not been comprehensively considered. Land sparing can be imperfect in two main ways: land not required for food production may not be used for conservation (incomplete area sparing), and habitat spared may be of lower quality than that assessed in surveys (lower habitat quality sparing). Here we use published data relating population density to landscape-level yield for birds and trees in Ghana (167 and 220 species respectively) and India (174 birds, 40 trees) to assess effects of imperfect land sparing on region-wide population sizes and hence population viabilities. We find that incomplete area and lower habitat quality imperfections both reduce the benefits of a land-sparing strategy. However, sparing still outperforms sharing whenever >= 28% of land that could be spared is devoted to conservation, or the quality of land spared is >= 29% of the value of that surveyed. Thresholds are even lower under alternative assumptions of how population viability relates to population size and for species with small global ranges, and remain low even when both imperfections co-occur. Comparison of these thresholds with empirical data on the likely real-world performance of land sparing suggests that reducing imperfections in its delivery would be highly beneficial. Nevertheless, given plausible relationships between population size and population viability, land sparing outperforms land sharing despite its imperfections. Policy implications. Our results confirm that real-world difficulties in implementing land sparing will have significant impacts on biodiversity. They also underscore the need for strategies which explicitly link yield increases to setting land aside for conservation, and for adoption of best practices when spared land requires restoration. However, land-sparing approaches to meeting human agricultural demand remain the least detrimental to biodiversity, even with current imperfections in implementation.