Jerusalem, 16 September, 2026 (TPS-IL) — Artificial intelligence models can identify patterns in animal sounds, but they may still struggle to determine what those sounds mean to the animals hearing them, according to a new international study led by Israeli researchers.
“Identifying acoustic patterns is not necessarily the same as deciphering meaning,” said Prof. Yossi Yovel of Tel Aviv University, one of the study’s researchers. “To understand what an animal is ‘saying,’ we need to know how the animal receiving the message perceives it and responds to it.”
The study suggests that relying on AI to analyze the physical characteristics of animal vocalizations could provide an incomplete picture of animal communication. Sounds that are acoustically similar may convey different messages, while sounds that sound different may carry the same meaning.
The researchers tested this problem using vocalizations made by human toddlers who had not yet fully developed speech. Unlike animal calls, the communicative context of these sounds can be assessed to some degree because researchers can observe the circumstances in which they occur and how people respond.
The recordings included vocalizations made in three situations: distress, calling specifically for the child’s mother or father, and requesting food.
The researchers analyzed the recordings using a traditional acoustic method and two advanced neural networks. One had been trained using animal vocalizations, while the other had been trained on adult human speech. The comparison allowed the researchers to examine whether models trained on different types of vocal communication could identify the distinctions between the toddlers’ vocalizations.
The neural networks performed better than the traditional acoustic method, but neither could consistently classify the vocalizations according to their communicative context.
In some cases, the AI models grouped together sounds associated with different situations. In others, they separated sounds that were associated with the same situation. The models also failed to detect increasing urgency across sequences of vocalizations, a distinction that humans can perceive naturally.
AI Struggles to Decode Animal Sounds
The findings raise questions about recent efforts to use AI to study communication among bats, whales, birds and other animals. According to the researchers, such efforts often assume that identifying patterns in the physical characteristics of sounds can reveal what those sounds communicate.
But the meaning of a vocalization depends not only on how it is produced, but also on how it is perceived by the animal receiving it.
The researchers said that different species have their own perceptual worlds, meaning that a sound’s significance cannot necessarily be inferred from its acoustic characteristics alone.
They argue that future attempts to decipher animal communication should combine AI with behavioral observations, playback experiments and, where possible, measurements of brain activity. Such methods could help researchers determine not only which sounds occur together, but also how animals interpret and respond to them.
“In recent years, there has been growing excitement about the possibility of using artificial intelligence to decode animal communication, but our study shows that these promises should be treated with caution,” Yovel said.
“Artificial intelligence is a powerful tool, but it is no substitute for the perspective of the animal itself,” he added.
The research team included scientists from Tel Aviv University, the Hebrew University of Jerusalem, the University of Edinburgh in Britain, Germany’s Museum für Naturkunde – Leibniz Institute for Evolution and Biodiversity Science, and Humboldt-Universität zu Berlin.
The researchers said the findings do not rule out AI as a tool for studying animal communication, but indicate that acoustic analysis alone may not be sufficient to establish what animals are communicating.
The findings were published in the peer-reviewed journal Current Biology.