Prior to the construction of an offshore wind farm at the Belgian Thorntonbank, local seabird abundance was studied by means of ship-based surveys. ‘Seabirds at sea’ count data, however, exhibit extreme spatial and temporal variation, impeding the detection of human impacts on seabird abundance and distribution. This paper proposes a transparent impact assessment method, following a before–after control–impact design and accounting for the statistical challenges inherent to ‘seabirds at sea’ data. By simulating a broad range of targeted scenarios based on empirical model coefficients, we tested its efficacy in terms of power and investigated how the chance of statistically detecting a change in numbers is affected by data characteristics, monitoring period and survey intensity. Because of high over-dispersion and/or zero inflation, the power to detect a 50% decrease in numbers was generally low, but did reach 90% within less than 10 years of post-impact monitoring for northern gannet (Morus bassanus) and common guillemot (Uria aalge).