Intro — May 2026, AI nature ID has reached "expert level"

Back in 2020, asking your phone to identify a bird species from a single call felt like magic. As of May 2026, it is one tap inside **Merlin Bird IDs Sound ID**. The free app, run by the Cornell Lab of Ornithology, reports species-level accuracy above 96% across the eastern US and above 85% across Eurasia, while iNaturalist has crossed 200M cumulative observations and built the citizen-science backbone that feeds global biodiversity research.

This piece is not a marketing matrix. It honestly maps "which app fills which slot on the ground today" — birds, plants, insects, mushrooms, rocks, plus hiking and trail apps, including Koreas MoYaMo and Kiwi Live and Japans YAMAP and YAMARECO. This is the 2026 lineup from end to end.

AI Nature ID 2026 — broken down into 7 categories

First, the big picture. The 2026 nature-observation app market splits into 7 categories.

1. **Bird ID — audio + photo**: Merlin Bird ID, BirdNET, Picture Bird

2. **Bird logging / citizen science**: eBird, iNaturalist

3. **General wildlife ID**: iNaturalist, Seek

4. **Plant ID**: Pl@ntNet, PictureThis, PlantSnap, MoYaMo

5. **Insect / mushroom / mineral ID**: Picture Insect, Picture Mushroom, Rock Identifier

6. **Hiking / trail / nature route apps**: AllTrails, Komoot, Gaia GPS, YAMAP, Tranggle

7. **Smart cameras / feeders**: Bird Buddy, Birdfy (Netvue)

Within each category, OSS / academic / SaaS / government-data tracks diverge. We walk through them one by one.

Merlin Bird ID — the free Cornell-Lab juggernaut

Cornell Lab of Ornithologys **Merlin Bird ID** is the undisputed leader of bird ID apps as of May 2026. Free, ad-free, open-data-backed — those three together are the moat.

It packs three ID modes into one app.

- **Sound ID**: streams the microphone in real time and infers species, with a visualised spectrogram. Accuracy has improved sharply every year since the 2021 launch; as of May 2026 it covers roughly 1,000+ species worldwide.

- **Photo ID**: take a photo and a vision model narrows candidates. eBird observation data acts as a location/seasonal prior.

- **Step-by-Step**: a five-question decision tree on colour, size, behaviour, and habitat. Non-AI branching, so it works offline.

Both the audio and vision pipelines are widely understood to use **CNN backbones with transformer post-processing**. Training data is Cornells Macaulay Library audio plus eBird photos and location metadata.

Local example using Cornells academic OSS BirdNET-Analyzer.

Merlin itself is closed, but the same lab releases BirdNET as OSS.

pip install birdnetlib

python - <<'PY'

from birdnetlib import Recording

from birdnetlib.analyzer import Analyzer

analyzer = Analyzer()

rec = Recording(

analyzer,

"field-recording.wav",

lat=37.5665, lon=126.9780, # Seoul coordinates

date=__import__("datetime").datetime(2026, 5, 16),

min_conf=0.25,

)

rec.analyze()

for d in rec.detections:

print(d["common_name"], d["confidence"])

PY

Merlin is a **self-contained citizen-science tool**. Korean and Japanese users download the **Northeast Asia** region pack, which covers Korean, Japanese, and East Sea coast species in a single bundle.

BirdNET — the OSS baseline for acoustic ML

**BirdNET** is the acoustic bird-ID model jointly developed by Cornell Lab and TU Chemnitz. If Merlin is the closed-source consumer app, BirdNET is the **academic / OSS backend**.

- **BirdNET-Analyzer**: CLI plus Python library. Covers 6,000+ species.

- **BirdNET-Pi**: Raspberry Pi with a microphone for 24/7 backyard recording and identification — wildly popular for backyard monitoring.

- **BirdNET-Go**: a Go port of the same model for embedded / edge deployments.

Academic uptake is strong. As of May 2026 Google Scholar shows 6,000+ citations, and government monitoring programmes — Koreas Ministry of Environment / NIBR, Japans Ministry of the Environment, and others — use it as a backend.

24/7 garden monitoring with BirdNET-Analyzer

from birdnetlib import Recording

from birdnetlib.watcher import DirectoryMultiProcessingAnalyzer

from birdnetlib.analyzer import Analyzer

analyzer = Analyzer()

watcher = DirectoryMultiProcessingAnalyzer(

"/data/recordings",

analyzers=[analyzer],

lat=37.5665,

lon=126.9780,

min_conf=0.40,

)

watcher.watch()

Treat Merlin and BirdNET as **the same family with different interfaces**. End-users pick Merlin; researchers and operators running citizen-monitoring infrastructure pick BirdNET.

eBird — the global standard for citizen-science bird logging

**eBird** is also run by Cornell Lab. Launched in 2002, it now holds 1.7B+ cumulative observations as of May 2026 — the global standard log of bird sightings.

- Records by **checklist** with time, location, species, and counts.

- **Hotspots**: famous birding sites that users register. Seouls Han River Park, Incheons Songdo tidal flats, and Japans Yatsu-higata are all on the map.

- **eBird API**: free with rate limits. Observations, species distributions, and hotspot data are served as JSON.

eBird is a first-class input for academic research. Macaulay Library audio, eBird photos, and eBird checklists effectively form Cornells unified data asset.

iNaturalist — Cal Academy + National Geographic for all-wildlife ID

For everything that is not a bird, the de facto standard is **iNaturalist**. Co-run by the California Academy of Sciences and the National Geographic Society, the platform reports **200M+ cumulative observations and 3.2M+ active users** as of May 2026.

Two values matter most.

1. **AI first-pass ID, then expert / community verification**: when users upload a photo, the vision model proposes candidate species; when the community reaches consensus (Research Grade), the record flows to GBIF.

2. **Data flowing back to academia**: Research Grade observations are piped automatically to GBIF (Global Biodiversity Information Facility). Citizen science to global academic data, end to end.

The iNaturalist API is OAuth2-based and free; Korean and Japanese vernacular-name mapping has improved markedly since 2024.

iNaturalist Observations API example — Research Grade within 50km of Seoul, May 2026

resp = requests.get(

"https://api.inaturalist.org/v1/observations",

params={

"lat": 37.5665,

"lng": 126.9780,

"radius": 50,

"month": 5,

"year": 2026,

"quality_grade": "research",

"per_page": 50,

},

timeout=10,

)

data = resp.json()

for obs in data["results"][:10]:

name = obs.get("taxon", {}).get("name")

place = obs.get("place_guess")

print(name, "@", place)

The killer feature of iNaturalist is the **community correction loop** — even when the AI is wrong, expert volunteers catch it. That is the biggest gap versus apps that trust a single model.

Seek by iNaturalist — gamified for kids and beginners

Built by the same team, **Seek** is the family- and elementary-school-friendly simplification of iNaturalist.

- No account required; point the camera at a plant, insect, or animal and the species ID surfaces progressively in real time.

- Observations are stored anonymously and a **Field Guide** fills out automatically.

- Location data is heavily anonymised, which makes it safe for school and family use.

The underlying model is a mobile build distilled from iNaturalists vision model. As of 2026, its accuracy beats the average non-specialists ID ability.

Audubon Bird Guide & iBird Pro — the US field-guide lineup

The US market has the Audubon Societys **Audubon Bird Guide** plus the commercial **iBird Pro** field guide.

- **Audubon Bird Guide**: free. 800+ species, offline operation, audio recordings, seasonal guides.

- **iBird Pro**: paid. Ultra-detailed illustrations, comparison search, notes.

Merlin handles AI first-pass ID, eBird handles logging, and Audubon / iBird handle the "guidebook + reading material" role. US birders typically install all three.

Sibley Birds & Collins Bird Guide — digital field-guide canon

Authoritative paper field guides have been digitised cleanly.

- **Sibley Birds 2nd Edition**: David Sibleys illustrated guide as an app. 800 North American species.

- **Collins Bird Guide**: the de facto standard for European / UK birding, available on iOS and Android.

- **Birds of Korea / NIBR Bird DB**: the National Institute of Biological Resources runs Koreas bird and wildlife DB.

This lineup is about **authoritative illustrations and audio**, not AI inference. The flow is to ID with an AI tool, then read more in a field guide.

Pl@ntNet — the academic baseline for plant ID

The academic standard for plant ID is **Pl@ntNet**, jointly run by CIRAD / INRA / IRD / INRIA in France, with support from the Agropolis Fondation.

- Free, ad-free, open-data.

- Covers 40,000+ species as of May 2026.

- An active-learning loop adds new species to the model once enough citizen photos have been collected.

Pl@ntNets raw accuracy is sometimes called slightly behind commercial apps such as PictureThis, but its **public data release** is the decisive feature for academics and ministries.

PictureThis, PlantSnap, NatureID — commercial plant ID

The consumer market is dominated by commercial plant ID apps.

- **PictureThis**: #1 on both accuracy and UX. Subscription model. Adds gardening guides and disease diagnosis.

- **PlantSnap**: strong in the US market. Free with ads.

- **NatureID**: an all-in-one plant + insect + bird/animal ID app.

Their business model wraps **AI ID into a gardening / plant-care / disease-diagnosis subscription** — clearly a different track from free academic tools such as Pl@ntNet.

Google Lens & LeafSnap — general-purpose vision in the wild

You do not need a dedicated app. **Google Lens** is a routine first-pass tool for nature ID. There is no separate plant/animal mode, but Lenss general vision model is accurate enough for common species.

**LeafSnap** is a leaf-based ID app co-developed by Smithsonian and Columbia, with leaf-shape training data as its strength.

The tradeoff between general and specialist is clear. Lens is best for "what even is this?" first-pass screening; Pl@ntNet and PictureThis are stronger at narrowing to a specific species.

Picture Insect & Insect ID — bug ID

Insect ID is much harder to model than birds or plants. The species count is overwhelming, and a single photo often does not narrow you to a species.

- **Picture Insect**: effectively #1 in the insect ID category. iOS and Android.

- **Insect ID**: a competing app in the same category. Subscription model.

- **Seek + iNaturalist**: insects via the same vision model. Accuracy varies by case.

Professional entomologists do not trust AI alone for species-level ID. Narrowing to **genus** and having a human expert finish is still the standard pattern.

Picture Mushroom & Shroomify — danger and opportunity in fungi ID

Mushroom ID is a category where **insufficient AI accuracy can cost lives**, since edible and toxic species often look alike.

- **Picture Mushroom**: #1 in the mushroom category, with an explicit "do not eat based on this app alone" warning in the UI.

- **Shroomify**: stronger in the UK / Europe, same warning policy.

- **Mushroom Observer (with iNaturalist integration)**: a citizen-science / community ID platform.

The biggest value — and risk — of this category is **UX that stops users from making "is this edible?" decisions with the AI**. Since 2024-2026, every major app forces a step-one warning that the app alone must never decide what to eat.

Rock Identifier & PictureMinerals — mineral ID

Minerals and rocks are intrinsically hard to ID from images alone, but **Rock Identifier** has emerged as the de facto leader.

- **Rock Identifier**: minerals plus gemstone ID. Subscription model.

- **PictureMinerals**: a similar positioning, second mover.

Crystal structure, hardness, and density — non-visual traits — are decisive for minerals, and photos hit a ceiling fast. This is the category where **human experts add the most value over AI**.

AllTrails, Komoot, Gaia GPS — the global hiking / trail trio

Nature observation eventually means going outside. The global big three for hiking and trail apps are these.

- **AllTrails (+ AI Trail)**: #1 in the US. User reviews, AI-recommended trails, offline maps (paid).

- **Komoot**: strong in Europe; cycling- and MTB-friendly.

- **Gaia GPS (Outside Inc.)**: backpacking / pro tier. Topo maps, satellite layers, hunting seasons, and many specialty layers.

- **FarOut (formerly Guthook)**: the #1 thru-hiking app for PCT / AT / CDT.

- **Hiking Project (REI)**: a free US option built on OSM data.

In 2026, **AllTrails AI Trail** has extended into trail recommendations, condition forecasts, and crowding predictions via an LLM-based stack. Komoot is moving along a similar line.

Korean hiking & nature apps — Sangyeongmap, Tranggle, MoYaMo, Kiwi Live

Heres the Korean lineup.

- **Tranggle**: #1 in Korean hiking. GPS logs, summit verification, route recommendations.

- **Sangyeongmap / Hiking Course**: Korean hiking-course DB.

- **MoYaMo**: #1 plant ID app in Korea; backed by Naver-affiliated capital. Strong Korean native plant + ornamental DB.

- **Flower Garden**: plant ID plus a gardening community.

- **Kiwi Live**: naturalist community for live observations and data sharing.

- **NIBR (National Institute of Biological Resources)** apps: government-run species DB for the Korean peninsula.

- **Jayeon Dream**: environmental / ecosystem monitoring app.

In particular, **MoYaMo** has overwhelmingly better Korean native-plant data than global apps. iNaturalist and Pl@ntNet sometimes miss Korean species; Korean users typically install **MoYaMo plus iNaturalist** together.

Japanese nature / hiking apps — YAMAP, YAMARECO, Ikimono-log

Japan has one of the most developed hiking / nature app markets in the world.

- **YAMAP**: #1 Japanese hiking app. GPS log, activity SNS, safety alerts.

- **YAMARECO**: hiking record / route / activity log community. Together with YAMAP, the dominant duo in Japan.

- **Ministry of the Environment "Ikimono-log"**: a government citizen-science platform.

- **Ornithological Society of Japan + birding-club apps**: academic and club-driven lineup.

- **PictureBird Japan**: Japanese-UI bird ID app.

- **Yamapp / PictHike**: hiking-companion and photo-logging apps.

YAMAP is so deeply rooted in Japan that "summit verification = a YAMAP log" is effectively standard. For AI nature ID, even in Japan, the leaders remain iNaturalist, Pl@ntNet, and Merlin rather than a YAMAP-native ID feature.

Bird Buddy & Birdfy — the rise of the smart bird feeder

Since 2023, the **smart bird feeder** category has grown fast.

- **Bird Buddy**: a Slovenian startup. AI camera plus automatic photo / video plus species ID — a game-changer for backyard birding.

- **Birdfy (Netvue)**: from Chinas Netvue; a strong Bird Buddy competitor with a price-competitive edge.

- **Wingscapes**: camera-trap / wildlife-camera lineup; pro / hunting / wildlife-monitoring tier.

This category combines **hardware + AI vision + cloud subscription**. ID accuracy is not at Merlin / iNaturalist level, but the constrained environment of a single garden and a fixed camera distance makes the user experience excellent.

GBIF & open biodiversity data — the academic data hub

The ultimate sink for nature-app data is **GBIF (Global Biodiversity Information Facility)**.

- The open hub for global biodiversity observations.

- Aggregates iNaturalist Research Grade records, eBird data, and museum specimen data.

- Cumulative observations have crossed 3B+ as of May 2026.

Researchers and environmental-policy analysts treat GBIF as a first-party data source. When a citizen uploads a photo through MoYaMo, Merlin, or iNaturalist, that data flows through GBIF and becomes an input to conservation policy.

AI tech — CNNs + transformers + audio spectrograms + geographic priors

The 2026 standard architecture for nature ID models looks like this.

- **Vision**: CNN backbones (EfficientNet family) with Vision Transformer post-processing. Multi-crop, multi-scale inputs.

- **Audio**: Mel spectrograms fed to CNNs as images. BirdNET is the canonical example.

- **Geographic prior**: location, season, and time-of-day metadata narrow the candidate distribution. eBird and iNaturalist data act as the prior.

- **Multimodal fusion**: photo + location + behaviour-text inputs are increasingly combined.

The relevant academic venues are the **NeurIPS Wildlife Conservation workshops**, **AAAI AI for Social Good**, and **CVPR FGVC (Fine-Grained Visual Categorization)**.

Citizen-science events — Project FeederWatch, Christmas Bird Count, BioBlitz

The citizen-science events that nature apps power are part of the story.

- **Project FeederWatch**: Cornell Lab plus Bird Studies Canada. Winter backyard bird counts.

- **Christmas Bird Count**: Audubon Society. A 100+ year tradition.

- **eBird Global Big Day**: a 24-hour global simultaneous count.

- **City Nature Challenge**: an iNaturalist-driven city-scale BioBlitz.

- **Korean NIBR Native Plant Monitoring**: run by the National Institute of Biological Resources.

The biggest change of the 2020s is that these events are essentially **unworkable without AI ID apps**. The implicit assumption that every participant is an expert has been dissolved by AI first-pass ID.

Conservation AI — Wildbook, AI for Earth, Rainforest Connection

Adjacent to nature ID is the **conservation-AI** lineup.

- **Wildbook (WildMe)**: individual ID of manta rays, zebras, and whales. CNNs match stripe and pattern features.

- **Microsoft AI for Earth**: a programme granting conservation NGOs cloud and AI credits.

- **NVIDIA Earth-2**: climate / environment simulation (covered separately).

- **Conservation Metrics**: automated acoustic monitoring with species inference.

- **Rainforest Connection**: refurbished phones, solar panels, and acoustic ML to detect illegal logging and poaching.

End-users do not install these, but they are the infrastructure that **connects citizen-science data into conservation policy**.

Korean government / academic lineup — National Parks, NIBR, NIER

A note on the Korean government / academic lineup.

- **Korea National Park Service**: trails, wildlife, and plant-conservation info. App and web.

- **National Institute of Biological Resources (NIBR)**: a species DB for the Korean peninsula, citizen monitoring, and AI ID research.

- **National Institute of Ecology**: ecosystem surveys, citizen science, environmental impact assessment.

- **National Institute of Environmental Research**: applied environmental conservation research.

NIBRs native-species DB acts as the baseline taxonomy that MoYaMo and Koreanised iNaturalist defer to.

Japanese government / academic lineup — MoE, KAHAKU, Yama-Kei

And the Japanese side.

- **Ministry of the Environment "Ikimono-log"**: governmental citizen-science data platform.

- **National Museum of Nature and Science (KAHAKU)**: Japans natural-history and species DB.

- **Yama-Kei Publishers (Yama to Keikoku)**: a magazine and field-guide publisher; many digital field guides.

- **Wild Bird Society of Japan**: the Japanese wild-bird conservation and citizen-science body.

Japans academic / publishing / government data are well connected, so the loop from citizen-science apps such as YAMAP and iNaturalist into government data is one step smoother than in Korea.

Stack patterns — how real nature observers combine apps

Finally, real user-persona stack patterns.

- **Beginner / family**: Seek + Merlin Bird ID + AllTrails free tier.

- **Korean hiker / botanist**: MoYaMo + Tranggle + Merlin Bird ID (Northeast Asia pack) + iNaturalist.

- **Japanese hiker / naturalist**: YAMAP + YAMARECO + iNaturalist + Merlin + Ikimono-log.

- **North American birder**: Merlin + eBird + Audubon + AllTrails / Gaia GPS.

- **European birder / naturalist**: Collins Bird Guide + Merlin + Pl@ntNet + Komoot.

- **Researcher / environmental NGO**: iNaturalist + GBIF + BirdNET-Analyzer + Wildbook.

- **Backyard / smart-feeder user**: Bird Buddy or Birdfy + Merlin + Project FeederWatch.

The common shape is a three-step pipeline: **AI first-pass ID, then academic / community verification, then return into government / academic data**. Trying to do it all in one app always leaves a gap.

Closing — May 2026: "AI ID is just a tool; observation is still human work"

The opening claim — that AI ID accuracy has reached expert level — has a counter-intuitive conclusion. **The more accurate AI gets, the more important human observation, logging, and conservation become**.

Merlin catching a bird call is just the start; you log on an eBird checklist, post photos on iNaturalist, the data flows to GBIF, and it becomes an input to conservation policy. Do not spend forever picking tools. **Merlin + iNaturalist + one regional app (MoYaMo / YAMAP)** covers 90% of users.

The remaining 10% is time and footsteps. In 2026, AI still cannot do that for you.

References

- Cornell Lab of Ornithology — Merlin Bird ID: https://merlin.allaboutbirds.org/

- Cornell Lab — eBird: https://ebird.org/

- BirdNET — Cornell + TU Chemnitz: https://birdnet.cornell.edu/

- BirdNET-Analyzer (OSS, GitHub): https://github.com/kahst/BirdNET-Analyzer

- BirdNET-Pi (OSS, GitHub): https://github.com/mcguirepr89/BirdNET-Pi

- iNaturalist: https://www.inaturalist.org/

- iNaturalist API docs: https://api.inaturalist.org/v1/docs/

- Seek by iNaturalist: https://www.inaturalist.org/pages/seek_app

- Pl@ntNet: https://plantnet.org/

- PictureThis: https://www.picturethisai.com/

- PlantSnap: https://www.plantsnap.com/

- Audubon Bird Guide: https://www.audubon.org/app

- Sibley Birds: https://www.sibleyguides.com/product/sibley-birds-app/

- Collins Bird Guide: https://www.collins.co.uk/products/collins-bird-guide-app

- Picture Insect: https://pictureinsectai.com/

- Picture Mushroom: https://picturemushroomai.com/

- Rock Identifier: https://rockidentifierai.com/

- AllTrails: https://www.alltrails.com/

- Komoot: https://www.komoot.com/

- Gaia GPS: https://www.gaiagps.com/

- FarOut Guides: https://faroutguides.com/

- MoYaMo: https://www.moyamo.co.kr/

- Tranggle: https://www.tranggle.com/

- Kiwi Live: https://kiwilive.kr/

- National Institute of Biological Resources (NIBR): https://species.nibr.go.kr/

- YAMAP: https://yamap.com/

- YAMARECO: https://www.yamareco.com/

- MoE Japan — Ikimono-log: https://ikilog.biodic.go.jp/

- National Museum of Nature and Science (KAHAKU): https://www.kahaku.go.jp/

- GBIF — Global Biodiversity Information Facility: https://www.gbif.org/

- Wildbook (WildMe): https://www.wildbook.org/

- Microsoft AI for Earth: https://www.microsoft.com/en-us/ai/ai-for-earth

- Rainforest Connection: https://rfcx.org/

- Bird Buddy: https://mybirdbuddy.com/

- Birdfy (Netvue): https://www.netvue.com/pages/birdfy