AI-generated illustration created with Exas: an imagined person using a smartphone beside a woodland path; not a photograph of BirdNET users or a field test.

A woodland can be full of sounds before we have names for any of them. A call from the canopy, a brief trill from a hidden branch, a voice heard once and then lost: listening is often the beginning of a nature question.

AI sound analysis offers one way to explore those questions. The interesting story is not only the software. It is the collaboration between people who know conservation, people who build tools and people willing to pay attention to the living world.

A tool shaped around people in the field

On 9 September 2026, the Cornell Lab of Ornithology announced BirdNET Live, an app that runs wildlife-sound identification locally on a smartphone. The announcement describes its use by researchers, rangers and volunteers, including in places without a mobile connection. It was developed primarily for conservation professionals, researchers and students. [1]

Cornell identifies Stefan Kahl as BirdNET’s lead developer and describes cooperation between conservation practitioners and AI developers in shaping the app. That connection matters: a useful conservation tool needs to fit the questions people are trying to answer in the field. [1]

From a recording to a suggestion

BirdNET’s project documentation describes a sound-analysis pipeline that divides recordings into short segments, turns sound into spectrograms and uses a neural network to identify patterns. Location and date information can help refine the expected species. It is a technical way of finding possible identities in a recording, rather than direct access to what an animal is thinking. [2]

BirdNET Live stores recordings and associated time and location information on the phone. Cornell describes workflows for reviewing, exporting and sharing recordings with experts for validation. Keeping that review step in view helps separate a useful suggestion from a confirmed observation. [1]

The human part does not disappear

BirdNET’s research collection includes guidance on interpreting detector scores and work examining how performance changes with species, thresholds and recording context. A confident-looking output should therefore be considered alongside the recording and the conditions in which it was made. [3]

For someone beginning to notice birds, an uncertain result can become a reason to listen again or ask a knowledgeable person. For a monitoring project, decisions about recording methods and validation need more care than accepting every label the software produces. More detections do not, by themselves, establish that a habitat is healthy or improving.

Different ways to contribute

BirdNET’s contribution page describes several roles: developers can improve code and documentation; people with recordings can contribute through appropriate data-sharing routes; and regional experts can help validate detections. It also invites collaboration on conservation monitoring. The project’s needs extend beyond writing software. [4]

That makes a useful connection for a community such as Earth Love. Some people enjoy practical habitat work, others enjoy careful observation, and others enjoy technology. A shared question about nature can give those interests a reason to meet.

Begin with attention, then choose the tool

A modest first exercise is to spend a few quiet minutes listening from a suitable path, note what you hear and look for a way to learn more. If you try an identification tool, read its current instructions and keep uncertain results labelled as uncertain. This article is not a hands-on review of BirdNET Live.

Take care with recordings and location information before sharing them. Avoid including other people’s conversations, and think about whether a precise wildlife location should be public. Let the observation remain quiet and unobtrusive rather than trying to make an animal perform for a recording.

AI can help organise the sounds we notice. The responsibility to observe carefully, question a result and decide what to do next remains with people. For another view of shared care, read the people bringing Sweden’s meadows back to life.

Sources & further reading

Prepared with AI assistance and checked against the project documentation and Cornell announcement linked below. Earth Love has not field-tested BirdNET Live. The cover is an imagined scene, not a product demonstration or photograph of BirdNET users.