Aerial surveys using direct counts of animals are commonly used to estimate deer abundance. Forward-looking infrared (FLIR) technology is increasingly replacing traditional methods such as visual observation from helicopters. Our goals were to compare fixed-wing FLIR and visual, helicopter-based counts in terms of relative bias, influence of snow cover and cost. We surveyed five plots: four 41.4 km² plots with free-ranging white-tailed deer Odocoileus virginianus populations in Wisconsin and a 5.3 km² plot with a white-tailed deer population contained by a high fence in Michigan. We surveyed plots using both fixed-wing FLIR and helicopters, both with snow cover and without snow. None of the methods counted more deer than the other when snow was present. Helicopter counts were lower in the absence of snow, but lack of snow cover did not apparently affect FLIR. Group sizes of observed deer were similar regardless of survey method or season. We found that FLIR counts were generally precise (CV = 0.089) when two or three replicate surveys were conducted within a few hours. However, at the plot level, FLIR counts differed greatly between seasons, suggesting that detection rates vary over larger time scales. Fixed-wing FLIR was more costly than visual observers in helicopters and was more restrictive in terms of acceptable survey conditions. Further research is needed to understand what factors influence the detection of deer during FLIR surveys.