Brno, 21 November 2025
Conspecific brood parasitism – a situation where females lay eggs in the nests of other females of the same species – has long been a mystery to scientists. A team from the Institute of Vertebrate Biology of the Czech Academy of Sciences, in collaboration with colleagues from the UK, has now developed an artificial intelligence model that can recognise parasitic eggs in birds with up to 97% accuracy. The new approach, the results of which were published in the journal Proceedings of the Royal Society B: Biological Sciences, could significantly facilitate and reduce the cost of research into this little-studied phenomenon.
Some birds use so-called brood parasitism to reproduce – they lay their eggs in the nests of other birds and leave the care of their young to the new ‘foster parents’. The best-known example is the common cuckoo (Cuculus canorus), which parasitises other bird species. However, conspecific brood parasitism is much more common, although less researched. This phenomenon has been observed in some ducks (e.g. common pochards and goldeneyes), but also in house sparrows and European starlings.
While cuckoo eggs often differ from host eggs at first glance, it is much more difficult to detect parasitic eggs in the nest of the same species because they differ only minimally from the other eggs already in the nest. A new study has shown that even experienced ornithologists often make mistakes when visually identifying parasitic eggs. Previous research has therefore relied primarily on genetic testing, which provides reliable results but is costly and time-consuming, and is one of the reasons why conspecific brood parasitism remains a relatively unexplored behaviour.
Artificial intelligence helps detect parasitic eggs in birds
A team of scientists from the Institute of Vertebrate Biology of the Czech Academy of Sciences (ÚBO AV ČR), in collaboration with their English colleagues from the University of Essex and University of Exeter, has developed a machine learning model that can identify parasitic eggs based solely on photographs. The model was tested on the clutches of the barn swallow (Hirundo rustica), in which conspecific parasitism has been previously described. “To create it, we used data on the appearance of 270 swallow eggs from 54 different females and took into account all the main characteristics of the eggs – size, shape, colour and pattern,” explains Michal Šulc, lead author of the study from ÚBO AV ČR. “Based on this information, the model first learned to distinguish how similar eggs laid by one female are and how eggs differ between different females. We then created nearly 2,000 ‘artificially parasitised’ nests to test the model’s accuracy in identifying foreign eggs”, he adds. The result was extraordinary – the accuracy of artificial intelligence reached 97%, while humans (including experts) correctly identified parasitic eggs in 87% of cases on average.
The model developed in the study, along with detailed instructions for its use, are freely available and can greatly facilitate future research on brood parasitism in birds. The study thus shows that artificial intelligence can be an effective tool for detecting ‘bird cheats’ and contribute to a better understanding of reproductive strategies in the bird kingdom.
Link to the publication: https://royalsocietypublishing.org/doi/10.1098/rspb.2025.2085
