Grazing by domestic livestock has greatly degraded many Australian ecosystems and its legacy will be long-lasting in many areas. While livestock are usually removed from conservation reserves because they are perceived to be incompatible with the conservation of natural ecosystems, they have been retained in a number of reserves in south-eastern Australia as a management tool to achieve conservation outcomes. These cases highlight the fact that no framework currently exists to address the question: under what circumstances (and in what ecosystems) is livestock grazing – or the removal of grazing – likely to have positive, negative, neutral or uncertain impacts on native plant diversity and composition? This paper provides a conceptual framework to predict the effects of livestock grazing and grazing exclusion on the conservation values of native vegetation across natural ecosystems in Australia. It should prove equally relevant to other ecosystems around the world which evolved without heavy grazing by large herbivores. The framework is based on disturbance and grazing ecology literature from Australia and elsewhere, and incorporates six main factors: (1) impacts of livestock grazing on soil and ecosystem processes, (2) historical exposure to grazing, (3) site productivity, (4) relative palatability of dominant species, (5) species-specific factors influencing plant recruitment, and (6) spatial scale and landscape context. These factors are integrated into a decision tree to describe the potential impacts of livestock on native vegetation in a particular area. Livestock grazing is likely to have detrimental impacts on conservation values in many ecological contexts, especially in relatively intact, uninvaded ecosystems on unproductive soils. By contrast, it may be a useful management tool to achieve conservation objectives where it either: (1) controls the biomass of existing potentially-dominant, grazing-sensitive plants (native or exotic), (2) prevents encroachment by undesirable, grazing-sensitive, potential dominants, (3) provides disturbance niches required by rare or significant plant species, (4) maintains fauna habitat structure, or (5) enhances the diversity of species and vegetation structures across the landscape, especially when most of the landscape is ungrazed. In many cases, other disturbance regimes (especially burning) may achieve similar outcomes; however, other disturbances will not necessarily be more effective than grazing per se, especially in degraded or invaded areas. The framework provides a coarse-level filter to inform management decisions and to allow the findings from individual studies to be placed in a larger ecological context. While the framework is intended to improve decisions about conservation management, it is clear that much more research is needed to assess the role of grazing exclusion in previously grazed ecosystems, while modifications to current grazing regimes require testing, perhaps using adaptive management principles, to ensure optimal outcomes for biodiversity conservation.