Use of artificial nests in studies of nest success in birds has become increasingly popular, but this practice has been criticised because the appearance and success of artificial nests may differ from natural nests. Typically, artificial nests are presented with few visible eggs throughout the entire exposure period while natural nests during their life have different appearances, possibly producing stage-dependent predation rates. In ducks, nests may appear with openly visible eggs, eggs covered with nest material and covered by an incubating female. To test how nest appearance may affect nest survival, I simulated such duck nest appearances, and used direct observations to record responses of marsh harriers Circus aeruginosus, the dominant duck nest predator at the Lake Engure, Latvia, during the breeding seasons 2000–2002. I recorded 274 harrier flights over all types of artificial nests, 74 nest discoveries and 54 predation events. Logistic regression analyses showed that nest appearance was a significant predictor of the fate of the nest: nests with openly visible eggs were discovered more often than nests covered with nest material. Nests with dummy females had discovery rates similar to the average of all artificial nest types suggesting that they best reflect the predation risk of artificial nests and could be used in future studies. After discovering nests covered with dummy females, harriers never attacked but tried to scare them off to get access to the eggs. In 52% of cases of nest discovery, harriers gave up before touching the dummy. This shows that duck passive nest defence may prevent clutch predation.