The Cornell Lab of Ornithology's Merlin bird ID app will feed real-time identifications directly into eBird, the Lab's own citizen-science database, which already holds more than 2 billion observation records. The eBird mobile app will gain the ability to upload recordings captured in Merlin — a pipeline that, for the first time, connects mass-market acoustic AI to the world's largest structured bird-observation repository.

The scale of the user base makes this operationally significant. Merlin has crossed 40 million downloads across 240 countries; as recently as December 2025 that figure stood at 33 million, meaning the app added roughly 7 million installs in roughly six months. Britain is the second-largest user base globally, with close to 2 million UK users recorded in May 2026 alone. Since 2021, Merlin has used machine learning to identify birds by sound, returning a species match alongside an image — a capability that lowered the technical barrier for casual observers to generate structured data.

eBird, founded by the Cornell Lab in 2002 and managed by its Center for Avian Population Studies, functions as the backbone of global avian monitoring. Researchers, conservationists, and policymakers draw on its records to track population trends at continental and intercontinental scales. Routing Merlin identifications into that repository in near real-time would materially expand the geographic and temporal density of eBird's coverage — particularly in regions where trained surveyors are sparse.

The Accuracy Question

The integration arrives with a methodological caveat that the scientific community has not papered over. The European Bird Census Council explicitly recommends against using Merlin in official breeding bird surveys. The EBCC has also established a monitoring group to coordinate acoustic bird monitoring standards across Europe — a structural response to the proliferation of automated ID tools and the inconsistent data quality they can introduce.

The concern is not purely theoretical. Prof. Richard Gregory of the RSPB noted publicly that Merlin identified his dachshund as a mallard — a pointed illustration of how ambient noise and non-avian sounds can defeat the classifier. As of July 2026, Merlin can identify 2,066 species, covering most birds in the US, Canada, and Europe, but coverage thins considerably in the tropics and parts of Asia where training data is sparser.

Jessie Barry, who leads the Merlin project at Cornell Lab as Program Manager, and eBird Project Leader Marshall J. Iliff are the principal figures managing this integration. The Lab has not, at least in public statements captured to date, specified what quality-filtering or metadata-flagging will be applied to Merlin-sourced records before they enter eBird's core dataset. That distinction — whether AI-generated records are pooled with observer-vetted data or held in a separate analytical tier — will determine how much scientific weight downstream studies can assign to them.

Stakes for Conservation

The broader stakes are not abstract. The British Trust for Ornithology estimates the UK bird population has fallen by more than 70 million individuals over the past 50 years. Globally, avian population declines are among the more reliable early indicators of ecosystem stress. The value of denser monitoring data — even imperfect data — is that it can surface local collapses before they register in decadal survey cycles.

The tension the Merlin-eBird integration crystallises is one familiar to any field that has absorbed large-scale automated sensing: volume versus verifiability. Traditional eBird records carry observer-level accountability; a human submitter can be queried, their record disputed, their site revisited. An ML-generated submission is reproducible only if the original audio is retained and the model version is logged. Whether Cornell Lab's architecture preserves that provenance chain at scale is the technical question that will determine whether this integration expands the scientific record or merely inflates it.

A free Cornell Lab account already works across both Merlin and eBird, which means the authentication layer is in place. The remaining engineering and governance choices — flagging, tiering, provenance logging — are where the scientific value of this integration will actually be decided.