Protected areas vary enormously in their contribution to conserving biodiversity, and the inefficiency of protected area systems is widely acknowledged. However, conservation plans focus overwhelmingly on adding new sites to current protected area estates. Here we show that the conservation performance of a protected area system can be radically improved, without extra expenditure, by replacing a small number of protected areas with new ones that achieve more for conservation. Replacing the least cost-effective 1% of Australia's 6,990 strictly protected areas could increase the number of vegetation types that have 15% or more of their original extent protected from 18 to 54, of a maximum possible of 58. Moreover, it increases markedly the area that can be protected, with no increase in overall spending. This new paradigm for protected area system expansion could yield huge improvements to global conservation at a time when competition for land is increasingly intense.