Biological collections are increasingly becoming databased and available for novel types of study. A possible limitation of these data, which has the potential to confound analyses based on them, is their biased composition due to non-random and opportunistic collecting efforts. While geographic biases are comparatively well studied and understood, very little attention has been directed at other potential biases. We used Asteraceae specimen data from Australia’s Virtual Herbarium to test for over- and under-representation of plants with specific morphology, phenology and status by comparing observed numbers of specimens against a null distribution of simulated collections. Strong collecting biases could be demonstrated against introduced plants, plants with green or brown inflorescences, and very small plants. Specimens belonging to species with very restricted areas of distribution were also found to be strongly underrepresented. A moderate bias was observed against plants flowering in summer. While spiny plants have been collected only about half as often as should be expected, much of this bias was due to nearly all of them also being introduced (thistles). When introduced species were analyzed alone, a negative effect of spines remained but was much more moderate. The effect of woody or herbaceous habit, other inflorescence colours, tall growth and size of the capitula was comparatively negligible. Our results indicate that care should be taken when relying on specimen databases or the herbaria themselves for studies examining phenology, resource availability for pollinators, or the distribution and abundance of exotic species, and that researchers should be aware of collecting biases against small and unattractively coloured plants.