There has been considerable recent concern about the plight of seabirds globally, as many species have declined substantially. In the UK there are statutory needs to monitor seabirds at sea, particularly in light of new offshore areas being designated for conservation and plans for major offshore wind farm developments. However, the extent to which at-sea surveys are capable of detecting changes in abundance and options for improving survey protocols have received little attention. We investigate the power of detecting changes in numbers using at-sea surveys. Using data collected as part of a visual aerial seabird survey programme that covered areas of ‘Round 2’ offshore wind farm developments in UK waters, we quantify the variability and characterize the statistical properties of count data. By generating random datasets with the same properties as real data, we estimated the power of being able to detect various declines (50, 33, 25, 15 and 10%) and assessed the effects of survey duration and frequency and of spatial scale and variability in bird numbers. The results indicate that the survey design protocols used for the UK ‘Round 2’ offshore wind farm visual aerial seabird survey programme do not provide adequate means of detecting changes in numbers, even when declines are in excess of 50% and assumptions regarding certainty are relaxed to < 80%. Extending the duration, frequency and spatial extent of surveys would increase the probability of detecting changes, but not to a desirable level (e.g. > 0.8). The primary reason why there is a low probability of being able to detect consistent directional changes is that seabird numbers fluctuate greatly at any given location. Means of explaining this fine-scale variability are required, especially if small changes in populations are to be detected. Incorporating hydrodynamic variables into trend analysis might increase the power of detecting changes. Failure to detect changes in seabird numbers should not be taken to mean that no changes are occurring.