Effective conservation requires reliable data on the abundance and distribution of animals in space and time. During ship-based or aerial surveys for diving marine vertebrates such as sea turtles and marine mammals, a proportion of animals in a surveyed area will be missed because they are diving and out of view. While it is likely that dive and surface times vary with environmental conditions, such variation is rarely incorporated into survey-based research and its consequences for analyses of survey data are not well known. We quantified the effects of neglecting to account for variation in the dive-surfacing patterns of green turtles (Chelonia mydas) and loggerhead turtles (Caretta caretta) when analyzing boat-based survey data from a foraging ground in Western Australia. We found that analyses of turtle sightings data can be confounded by variation in the probability of turtles being at the surface where they are available for detection. For example, during the cold season in deeper areas in Shark Bay, green and loggerhead turtle density was underestimated by 45% and 21%, respectively, if extended dive times relative to population medians were not accounted for. These results have important implications for applications of survey data for a variety of taxa including other sea turtles, marine mammals and large sharks that are surveyed by boat or plane. Diving and depth use studies have much to contribute to the assessment and management of these groups, which include many species of conservation concern.