AI Is Learning to Listen to Forests, Oceans, and the Birds in Your Backyard

Counting wildlife has always been slow, patient work. A biologist sits in a blind, a volunteer walks a transect with a clipboard, or a researcher spends evenings matching photos by eye. A growing set of AI tools is changing that, not by replacing the people who care about nature, but by helping them sort through the enormous amount of sound and imagery that modern sensors and phones now collect.

Several of those projects made news this fall, and one of the most useful versions of the technology is already sitting in many gardeners’ pockets.

A forest that listens back

On October 6, Venture Taranaki announced that two New Zealand conservation organizations, 800 Trust and Wildlife.ai, had been selected for Google DeepMind’s AI for the Planet Accelerator. Wildlife.ai’s Victor Anton described the cohort as only sixteen organizations, and both groups have just returned from a boot camp in Singapore that opened an intensive three-month program of technical help and mentorship.

800 Trust built Hark, an acoustic monitoring system that uses AI-powered sensors installed in remote forest. According to the announcement, the devices record and analyze bird calls, detect pest species, and track changes in the ecosystem over time. The trust is restoring thousands of acres of retired farmland in East Taranaki to native forest, so a tool that can tell whether birds are returning is a practical measure of whether the work is paying off.

Wildlife.ai is developing Wildlife Watcher, an open-source AI camera meant to identify and count smaller mammals, reptiles, and amphibians, animals that are typically harder to monitor with traditional methods.

“AI has enormous potential to make wildlife monitoring faster and more affordable, making the important work of conservation groups that much more impactful,” said Miranda Wells of 800 Trust.

Local knowledge plus machine learning

A similar idea is taking shape in the Peruvian Amazon. On October 1, Acaté Amazon Conservation announced SentaAI, a machine-learning tool that combines camera traps and bioacoustic recorders with the ecological expertise of the Matsés people in their ancestral rainforest. Senta is the Matsés name for the endangered red uakari monkey.

The organization says satellites are good at spotting lost forest cover but reveal little about what is happening beneath the canopy. The Matsés, who recognize more than 47 types of rainforest habitat, choose where to place the cameras and recorders to capture rarely seen species. SentaAI then processes the flood of images using a mix of publicly available recognition systems and a custom workflow developed over two years. The group says it is now training the tool to identify bird, amphibian, and mammal calls.

That pairing matters. The software is fast, but it depends on people who know where to listen in the first place.

A photo album for whales

At sea, the same approach has already proved itself. Happywhale, a platform founded in 2015, identifies individual whales by matching the unique markings on flukes and fins against every photo in its database. In a profile published this week, Forbes reported that the platform now holds more than 1.6 million photos uploaded by 53,000 people from more than 100 countries, with a 97 to 99 percent accuracy rate. In July it won the 2026 AI for Good Impact Award in the AI for planet category.

Co-founder Ted Cheeseman told Forbes the system delivered “a thousand-fold increase in efficiency” in identifying animals. Its records have informed more than 40 peer-reviewed studies, including research documenting the loss of roughly a third of Hawai‘i’s humpback population during the 2013 to 2016 marine heatwave. Much of that data came from tourists who simply uploaded photos from whale-watching trips.

Cheeseman and colleagues also offered a useful caution in a recent opinion piece: conservation success depends on scientists approaching AI “as captains rather than crew,” meaning people still have to design, interpret, and oversee these systems.

The version in your pocket

Backyard birders do not need a research grant to try this kind of tool. The Cornell Lab of Ornithology’s free Merlin Bird ID app includes a feature called Sound ID that listens to birds around you and shows real-time suggestions of who is singing.

Cornell has explained how it works. As the phone records, Merlin turns the audio into a spectrogram, an image of sound frequencies over time, and feeds that image into a deep convolutional neural network, the same family of models used to recognize objects in photos. Experts from the Macaulay Library and the eBird community trained the original model by marking the exact moments each bird was vocalizing in recordings, which helps the system separate a cactus wren’s rattle from a passing car. The lab says Sound ID runs on the device without needing a network connection.

Sound ID now covers 2,066 species, including comprehensive coverage of the United States and Canada, and Cornell updates it twice a year. Its list for the region includes familiar desert voices such as Gambel’s quail, cactus wren, Gila woodpecker, curve-billed thrasher, verdin, and Abert’s towhee. Cornell suggests comparing each recording with the songs and calls stored in the app to confirm what you heard, and people can help improve the model by uploading recordings to eBird checklists.

Where to try it near the East Valley

Boyce Thompson Arboretum in Superior is an easy place to practice. The arboretum has been named an Important Bird Area by the National Audubon Society, and from October 2026 through April 2027 it offers a beginner-friendly guided bird walk every Wednesday from 8:15 to 9:30 a.m., except December 23. Walks are free with admission, limited to 15 people, and filled first come, first served at the Welcome Center. Loaner binoculars are available at the admissions window. The arboretum is open daily from 8 a.m. to 5 p.m. through April.

A guided walk is a good way to test an app honestly. When Merlin suggests a species, the docent can confirm it or point out what the app missed, and over a few weeks your own ear starts to catch up.

What gardeners can take from it

For anyone who plants for wildlife, listening tools offer a simple way to see whether a garden is working. Run Sound ID for a few minutes on a fall morning before adding native shrubs or a water source, then try again in spring. The list will not be perfect, but it gives a rough, repeatable record of who is visiting, the same kind of before-and-after measure that 800 Trust is building for an entire forest.

The technology is impressive, but every project here depends on people: Matsés experts placing recorders, tourists uploading whale photos, and volunteers tagging their own recordings. AI is making it much faster to hear what is out there. Deciding what to protect, and planting the habitat to do it, is still up to us.