It has been suggested that genetically modified herbicide–tolerant crops may benefit biodiversity because spraying of crops may be delayed until later in the growing season, allowing weeds to grow during the early part of the year. This provides an enhanced resource for arthropods, and potentially benefits birds that feed on these. Thus, this technology could enhance biodiversity. Using a review of weed phenologies and a population model, we show that many weeds are unlikely to benefit because spraying is generally delayed insufficiently late in the season to allow most to set seed. The positive effects on biodiversity observed in trials lasting one or two seasons are thus likely to be transient. For one weed of particular significance (Chenopodium album, fat hen) we show that it is unlikely that the positive effects observed could be maintained by inputs of seed during other parts of the rotation. However, we find preliminary evidence that if spraying can be ceased earlier in the season, then a viable population of late–emerging weeds could be maintained. This strategy could benefit weeds in both genetically modified (GM) and non–GM crops, but would probably lead to reduced inputs in GM systems compared with conventional ones.