ARTIFICIAL intelligence could help breeders detect problems earlier, but marketing claims cannot replace scientific validation, Dr Jan Govaere told the WBFSH General Assembly.

The Ghent University veterinarian explained how cameras and wearable sensors can identify behavioural changes preceding foaling or discomfort. However, experienced horse people remain essential to interpreting these signals.

Context matters: rolling after exercise may be normal, while similar behaviour overnight could indicate discomfort. Detecting movement does not establish its significance.

Commercial systems vary in reliability. Sensor contact, lighting, coat colour and camera position can affect results. Reliable technology requires diverse training examples, expert labelling and practical testing.

Govaere warned that high-accuracy figures can conceal frequent false alarms when detecting rare events. He also highlighted anatomically inaccurate AI-generated illustrations that risk spreading misinformation.

His message was that technology should support human expertise: “Let the AI do the counting, and let horse people do the thinking.”