Few beetle species can be detected with 95% confidence using pitfall traps

False absences in wildlife surveys make it difficult to identify metapopulation processes, increase uncertainty of management decisions and bias parameter estimates in habitat models. Despite these risks, the number of species that can be detected with a certain probability in a community survey has rarely been examined. I sampled beetles over 5 months using pitfall trap grids at three rainforest locations in Tasmania, Australia. I compared detection probability for dispersed and clustered sampling schemes using a zero-inflated binomial model and a simpler occurrence method to calculate the probability of detection. After excluding extremely rare species, I analysed 12 of 121 species. Only three to six species could be detected with 95% probability using a sampling effort that is frequently applied in ecological studies. A majority of common species had a mid summer peak in detection probability meaning that survey effort could be reduced from 5 to 2 months with only a small reduction in data quality. Most species occurred at only a proportion of sample points within locations. Despite the implied spatial structuring, three small grids within a location detected 10 of 12 species as effectively as large, dispersed grids. This study warns that as little as 5% of the beetle fauna may have a 95% probability of detection using the frequently applied pitfall trap method, highlighting a substantial limitation in our ability to accurately map the distributions of ground invertebrates. Whether very large sample sizes can overcome this limitation remains to be examined.