A crunch, a snap, a grind. The sound of an animal chewing may be the key to understanding what it eats, and researchers have now trained artificial intelligence to decode those sounds. A new study published in the journal Ecological Informatics reveals that AI can identify the type of food an animal is consuming simply by listening to the noise its jaws make. The finding opens a noninvasive window into the diets of wild creatures, from fish to mammals, without needing to capture or dissect them.
How crunching data became a menu
Scientists from the University of California, Santa Barbara and the University of Washington led the research. They recorded the chewing sounds of 11 animal species in controlled settings, including rays, turtles, and a marine iguana. Each animal was offered different foods such as squid, shrimp, and clams. The team then fed those audio recordings into a machine learning model. The AI learned to distinguish between food types based on the acoustic signature of each bite. The model achieved high accuracy, correctly identifying the food item in most cases.
Why local researchers and conservationists took notice
The study took place in the United States, but its implications reach far beyond. Understanding what predators eat is essential for managing ecosystems and protecting endangered species. Traditional methods often involve stomach content analysis, which requires killing the animal, or direct observation, which is time consuming and often impossible in murky water or dense forest. Acoustic monitoring, by contrast, is passive and continuous. For marine biologists studying rays or sea turtles off the coast of California, this technique could reveal shifts in prey availability linked to climate change or overfishing. Local conservation groups see it as a way to track the health of food webs without disturbing the animals.
From laboratory crunch to field recordings
The researchers acknowledge that the current model was trained in a controlled environment. Background noise, water movement, and multiple animals feeding at once could complicate real world use. Still, they are optimistic. The next step is to test the system in the wild, attaching underwater microphones to reefs or deploying audio recorders in terrestrial habitats. If successful, the approach could become a standard tool for ecologists. It would allow them to monitor diet remotely, over long periods, and across species that are difficult to study otherwise.
This study does not claim that AI can replace traditional fieldwork. It does, however, suggest that the everyday act of eating produces a wealth of data we have only just begun to hear. The crunch of a shell or the tear of flesh carries information, and machines can now translate that noise into knowledge. For researchers trying to understand the hidden lives of animals, that is a sound worth listening to.
