A common error in many studies is that a species may go undetected when they are actually present, thus leading to underestimates of the true population. Because imperfect detection can affect population estimates, we calculated detection probabilities of two rare bat species, Rafinesque's big-eared bats (Corynorhinus rafinesquii; RBEB) and southeastern myotis (Myotis austroriparius; SEM), when roosting in tree cavities. We used two methods: repeated surveys of the same tree cavity and a removal method where a tree cavity was eliminated from future surveys after a bat was detected. When repeated surveys were conducted, detection probabilities were 95% for each observer and 100% for two observers combined (n  =  43). Using the removal method, each observer independently detected bats in tree cavities in 92% of inspections; whereas, combined observations of two surveyors improved detection probability to 99% (n  =  27). We also evaluated count error for species identification and abundance using a two-observer method and compared it to digital infra-red imagery. Count errors were typically <4% and 38% when ≤20 bats and >20 bats were present, respectively, compared to digital images. When count errors occurred, most (64–73%) underestimated the number of bats present. Observers correctly identified species on 91% of occasions. Video recording tree cavities improved estimates of bat abundance and verified species composition. Repeated surveys and the removal method were similarly effective and suitable techniques for monitoring bat species occurrence; however, visual surveys were accurate for estimating abundance when number of bats was ≤20.