These startups may spot plant stress before humans can – Automated Home

A thirsty plant can make sounds before its leaves begin to show visible signs of stress. That idea sounds like science fiction, yet researchers have shown that plants can generate detectable acoustic or electrical signals when under pressure, and startups are turning those signals into tools for earlier crop alerts.

The breakthrough matters because plant stress is often easiest to fix before it becomes visible. If sensors and AI can read those early signals accurately, farmers could spot drought, disease, or pest pressure sooner and respond before small problems spread across an entire field.

Plants are making sounds humans cannot hear

Plants do not speak in the traditional sense, but they can emit ultrasonic sounds when exposed to environmental stress. Researchers have found that stressed plants generate clicks and popping noises in frequencies ranging from 40 to 80 kilohertz.

These sounds are beyond the range of human hearing, which explains why they went unnoticed for so long. Specialized sensors can detect them, however, and recent advances in artificial intelligence are helping researchers interpret what those signals might mean.

The sounds are not evidence of plant consciousness. Scientists emphasize that plants are not intentionally communicating. Instead, the noises are physical byproducts of changes occurring within the plant as it responds to stress.

Why stressed plants make noise

The leading explanation involves a process called cavitation. When water inside a plant becomes difficult to transport because of drought or damage, tiny air bubbles can form within the plant’s vascular system.

As those bubbles form and burst, they may create ultrasonic clicks and pops. Researchers have compared the sound to popcorn popping or bubble wrap being squeezed, though the frequencies are much higher than humans can hear.

Different stress conditions appear to create different acoustic signatures. Drought stress may generate one pattern, while physical injury, such as stem damage, can create another, opening the door to automated stress identification systems.

The race to decode plant signals

Several startups are now attempting to transform plant signals into actionable data. Rather than waiting for crops to show visible damage, these companies want growers to respond before problems become obvious.

The promise is significant. Earlier detection can help farmers protect yields, reduce unnecessary chemical applications, and improve resource efficiency across large agricultural operations.

Two companies have emerged as notable players in this space. California-based InnerPlant focuses on genetically engineered crop signaling, while Swiss biotech company Vivent Biosignals concentrates on electrical signals generated naturally within plants.

How InnerPlant approaches the challenge

InnerPlant’s CropVoice platform takes a different path. Instead of relying primarily on ultrasonic sounds, the company uses engineered soybeans that produce detectable optical signals when specific fungal stress events occur.

The company’s InnerSoy sensors are designed to signal when fungal threats emerge. According to InnerPlant, CropVoice signals can be detected 4 to 6 weeks before stress is detectable by other scouting methods.

That early warning window could give growers additional time to make informed decisions. Rather than treating entire fields as a precaution, farmers may be able to target interventions where they are most likely to deliver meaningful results.

Little-known fact: InnerPlant’s CropVoice network covered 50,000 acres across the U.S. Midwest in 2025 and is scaling to more than 500,000 acres in 2026.

Building a network of living sensors

CropVoice functions through a network of engineered sensor plots deployed across agricultural areas. These plants effectively serve as biological monitoring stations, revealing crop health conditions.

When targeted stress events occur, the engineered plants generate detectable responses that feed into the broader monitoring platform. The resulting information is then translated into practical recommendations for growers and agronomists.

The approach highlights an emerging trend in agricultural technology. Instead of observing plants solely from the outside, companies are increasingly turning crops themselves into sources of real-time environmental intelligence.

Vivent listens from inside the plant

Vivent Biosignals uses a different strategy. Rather than modifying crops genetically, the company attaches smart biosensors directly to plants and measures naturally occurring electrical activity.

Plants continuously generate tiny electrophysiological signals as they interact with their environment. Vivent’s technology captures those signals and analyzes them using machine learning systems designed to identify stress patterns.

The company’s PhytlSigns crop diagnostics platform aims to detect pathogens, pests, drought conditions, and other threats before visible symptoms appear. Continuous monitoring provides a stream of data that can support proactive decision-making.

Why electrical signals matter

One advantage of electrophysiological monitoring is that it effectively listens from within the plant. Instead of observing external symptoms, the technology captures biological responses as they occur.

Machine learning algorithms process incoming data in real time, helping distinguish between different causes of stress. That distinction is important because drought, disease, and pest pressure often require very different responses.

For growers, knowing not only that a plant is stressed but also why it is stressed could significantly improve intervention strategies and reduce unnecessary treatments across agricultural operations.

Interesting fact: A stressed plant can be “noisy” even when it looks silent to us. In a 2023 Cell study, researchers found that stressed tomato and tobacco plants emitted ultrasonic airborne sounds that could be recorded from a distance and classified by stress type, such as dehydration or cutting.

Artificial intelligence is making it practical

The growing interest in plant sensing depends heavily on advances in artificial intelligence. Raw acoustic and electrical signals are complex and difficult to interpret without automated analysis.

Researchers at Universidad Autónoma de Madrid demonstrated this potential using deep learning systems trained to classify plant stress. Their work combined spectral analysis techniques with convolutional neural networks based on a ResNet-50 architecture.

The resulting models achieved multiclass classification accuracy of roughly 79% while reaching 92% accuracy in a one-vs-rest verification system. Those results suggest AI can identify meaningful patterns hidden within plant-generated signals.

Scientific validation continues to grow

Interest in plant bioacoustics accelerated following research published in 2023 that demonstrated plants emit ultrasonic sounds when stressed. Those findings helped move the field from speculation toward measurable science.

Subsequent studies expanded the understanding of plant-generated signals and explored how artificial intelligence could classify stress conditions. Research has continued investigating whether different stressors produce consistently identifiable acoustic fingerprints.

Scientists are also exploring broader ecological implications. Evidence suggests that insects, bats, mice, and certain mammals can detect ultrasonic frequencies associated with plant stress, raising questions about how ecosystems respond to these signals.

TL;DR

• Researchers have shown that stressed plants emit ultrasonic sounds, creating opportunities for sensors and artificial intelligence systems to identify problems before visible symptoms appear.
• InnerPlant uses genetically engineered crops to generate stress signals that may provide warnings four to six weeks earlier than traditional field scouting methods.
• Vivent Biosignals monitors electrical activity inside plants, allowing machine learning systems to detect pathogens, pests, drought conditions, and other threats in real time.
• Artificial intelligence models are becoming increasingly effective at classifying plant stress signals, with some research achieving accuracy levels high enough for practical applications.

This article was made with AI assistance and human editing.

If you liked this, you might also like: