Summary1. As threatened species decline in number, they become more difficult to detect and assess, decreasing our ability to make informed conservation management decisions. Wildlife scat detection dogs are a promising non-invasive survey method for increasing detection rates of these species. However, a complexity of variables can reduce the certainty of this method, which may contribute to low uptake rates by land managers. One of the variables discussed in dog training literature is the potential for habitats with more complex vegetation to inhibit dispersion of target odours, reducing the dog's ability to detect scats. 2. We undertook the first experimental study to test the effects of habitat structure on scat detection dog performance. We used a domestic dog trained to detect scats from endangered spotted-tailed quolls to undertake 120 searches across 3 habitats in both winter and summer conditions in NSW, Australia. Scat searches were located in open grassland, woodland and dense heath, and we recorded temperature, relative humidity and wind speed. Performance was measured by recording the distance at which the scat was first detected, the total search duration and the success rate of detection in each habitat. 3. Scat detection rates were 83% or higher in all habitats, and there was no significant difference in first detection distances between habitat structures. However, within habitats there was a significant, positive relationship between first detection distance and total search duration in the most complex habitat (P = 0Á0002, R 2 = 0Á23, d.f. = 1, SE = 0Á53). 4.The results support other findings showing that detection dogs can work effectively across a diversity of habitats, but specifically demonstrate that searches in complex vegetation should allow for increased search effort compared to relatively open vegetation. Otherwise, the risks of a type II error, of failing to detect a species when it is present, are likely to be higher in these habitats and this has significant implications for conservation management. The high detection rates in our results are discussed in the context of the odour thresholds used to train wildlife scat detection dogs, and we recommend future documentation of training methods to allow comparison of surveys across species and sites.