Glyphosate is a key component of weed control strategies in Australia and worldwide. Despite widespread and frequent use, evolved resistance to glyphosate is rare. A herbicide resistance model, parameterized for Lolium rigidum has been used to perform a number of simulations to compare predicted rates of evolution of glyphosate resistance under past, present and projected future use strategies. In a 30-year wheat, lupin, wheat, oilseed rape crop rotation with minimum tillage (100% shallow depth soil disturbance at sowing) and annual use of glyphosate pre-sowing, L. rigidum control was sustainable with no predicted glyphosate resistance. When the crop establishment system was changed to annual no-tillage (15% soil disturbance at sowing), glyphosate resistance was predicted in 90% of populations, with resistance becoming apparent after between 10 and 18 years when sowing was delayed. Resistance was predicted in 20% of populations after 25–30 years with early sowing. Risks of glyphosate resistance could be reduced by rotating between no-tillage and minimum-tillage establishment systems, or by rotating between glyphosate and paraquat for pre-sowing weed control. The double knockdown strategy (sequential full rate applications of glyphosate and paraquat) reduced risks of glyphosate and paraquat resistance to <2%. Introduction of glyphosate-resistant oilseed rape significantly increased predicted risks of glyphosate resistance in no-tillage systems even when the double knockdown was practised. These increased risks could be offset by high crop sowing rates and weed seed collection at harvest. When no selective herbicides were available in wheat crops, the introduction of glyphosate-resistant oilseed rape necessitated a return to a minimum-tillage crop establishment system.