Outdoors & nature · sourced statistics
AI for Conservation Statistics 2026
This page collects figures on how artificial intelligence is used to monitor wildlife: camera traps with computer vision, acoustic recorders with sound classifiers, and satellite imagery for fishing and forests. It starts with the biodiversity baselines that these tools feed into, from WWF's Living Planet Index, the IUCN Red List, IPBES and the Protected Planet Report. Coverage is global. Most figures describe 2024 to 2026; a few older peer-reviewed results, such as the 2018 Snapshot Serengeti study and the 2019 IPBES assessment, are included because they remain the reference points in the field and are labeled by year.
Compiled by the FruitPunch research desk · Updated November 15, 2026
- figures
- 50
- sources
- 21
- publishers
- 15
The short answer
Monitored wildlife populations fell by an average of 73% between 1970 and 2020, according to WWF's Living Planet Index, and the IUCN Red List counted 48,646 threatened species out of 172,620 assessed in 2025. AI now does the first pass on the data used to track that decline: Wildlife Insights, led by Conservation International with Google, trains on over 35 million camera trap images across 1,295 species, and Google DeepMind's Perch helped one Hawaiian lab find endangered bird calls nearly 50 times faster. A 2024 Nature study used deep learning on satellite images to show that 72% to 76% of industrial fishing vessels are not publicly tracked.
Quote these
Key findings
Each finding names who published it. Copy it with the source attached, or jump to it in context.
- 73%#1
Monitored wildlife populations fell by an average of 73% between 1970 and 2020, according to WWF's Living Planet Report 2024.
Source: World Wildlife Fund, Oct 2024
In context - 48,646#2
The IUCN Red List included 172,620 species in October 2025, of which 48,646 were threatened with extinction.
Source: IUCN, Oct 2025
In context - 61%#3
61% of bird species worldwide have declining populations, up from 44% in 2016, according to the IUCN Red List update of October 2025.
Source: IUCN, Oct 2025
In context - 1,295#4
The Wildlife Insights AI model is trained on over 35 million labeled camera trap images covering 1,295 species (2026).
Source: Wildlife Insights, 2026
In context - 94%#5
Wildlife Insights models identify 94% of blank camera trap images with an error rate of less than 2%.
Source: Wildlife Insights, 2026
In context - 50x#6
Perch helped the University of Hawaii's LOHE Lab find endangered honeycreeper sounds nearly 50 times faster than its usual methods (2025).
Source: Google DeepMind, Aug 2025
In context - 72#7
72% to 76% of the world's industrial fishing vessels are not publicly tracked, according to a 2024 Nature study that used deep learning on satellite imagery from 2017 to 2021.
Source: Nature (Paolo et al.), Jan 2024
In context - 4.3 million#8
The world lost 4.3 million hectares of tropical primary forest in 2025, more than 11 football fields per minute, according to WRI's Global Forest Watch data.
Source: World Resources Institute, May 2026
In context
Reading the numbers
The case for AI in conservation starts with scale. WWF's Living Planet Report 2024 shows a 73% average decline in almost 35,000 monitored vertebrate populations since 1970, and IUCN's 2025 Red List counts 48,646 threatened species. Yet only a fraction of species are assessed, and IPBES puts the number at risk near 1 million. Monitoring at that scale produces far more images and audio than field teams can review by hand.
Camera traps are the most mature use. Wildlife Insights reports that blank frames can be up to 80% of a dataset and that its models filter 94% of them with under 2% error, while Google's open SpeciesNet was trained on more than 65 million images. Peer-reviewed results, from the 2018 Snapshot Serengeti work at over 93.8% accuracy to a 2025 Spanish study at a 96.2% F1 score, show that well-sampled species are now handled reliably. Rare species and under-sampled regions remain the weak point.
Bioacoustics is catching up fast. BirdNET lists coverage of more than 11,000 species, Google DeepMind's Perch 2.0 was trained on over 1.5 million labeled recordings, and the first Perch model passed 250,000 downloads. Rainforest Connection reports 7,025 species detected through its acoustic network. These are tool makers' own counts, which is why this page pairs them with independent baselines.
Satellites extend monitoring to places no sensor network reaches. A 2024 Nature study by Global Fishing Watch and partners matched 2 petabytes of imagery with 53 billion GPS positions and found 72% to 76% of industrial fishing vessels untracked. On land, WRI reports tropical primary forest loss of 4.3 million hectares in 2025, down 36% from 2024, while fires now account for about 43% of tree cover loss. AI makes these patterns visible; acting on them still depends on enforcement and funding.
The wildlife decline that monitoring has to measure
- 173%
Monitored wildlife populations fell by an average of 73% between 1970 and 2020, according to WWF's Living Planet Report 2024.
Source: World Wildlife Fund, Oct 2024
Exact wording at the source
Washington, DC (October 9, 2024) -There has been a catastrophic 73% decline in the average size of monitored wildlife populations* in just 50 years (1970-2020), according to World Wildlife Fund‘s (WWF) Living Planet Report 2024.
- 235,000
The Living Planet Index, run by the Zoological Society of London, tracks almost 35,000 vertebrate populations of 5,495 species from 1970 to 2020.
Source: World Wildlife Fund, Oct 2024
Exact wording at the source
The Living Planet Index, provided by the Zoological Society of London (ZSL), tracks almost 35,000 vertebrate populations of 5,495 species from 1970-2020.
- 385%
Freshwater wildlife populations showed the steepest average decline in the 2024 index, at 85%, ahead of terrestrial (69%) and marine (56%) populations.
Source: World Wildlife Fund, Oct 2024
Exact wording at the source
The steepest decline is in freshwater populations (85%), followed by terrestrial (69%) and then marine (56%).
- 495%
Monitored populations in Latin America and the Caribbean fell by an average of 95% from 1970 to 2020, the steepest regional drop in the Living Planet Index.
Source: World Wildlife Fund, Oct 2024
Exact wording at the source
WWF’s flagship Living Planet Report details sharp declines in monitored wildlife populations with the steepest drops recorded in Latin America and the Caribbean (95%), Africa (76%) and Asia–Pacific (60%), followed by North America (39%) and Europe and Central Asia (35%).
- 51 million
Around 1 million animal and plant species are threatened with extinction, many within decades, according to the 2019 IPBES global assessment, the most recent assessment of its kind.
Source: IPBES, May 2019
Exact wording at the source
The Report finds that around 1 million animal and plant species are now threatened with extinction, many within decades, more than ever before in human history.
- 620%
The average abundance of native species in most major land habitats has fallen by at least 20%, mostly since 1900, according to IPBES (2019).
Source: IPBES, May 2019
Exact wording at the source
The average abundance of native species in most major land-based habitats has fallen by at least 20%, mostly since 1900.
- 740%
More than 40% of amphibian species and almost 33% of reef-forming corals are threatened, according to the 2019 IPBES assessment.
Source: IPBES, May 2019
Exact wording at the source
More than 40% of amphibian species, almost 33% of reef forming corals and more than a third of all marine mammals are threatened.
- 82 million
More than 2 million species are known to science, with millions more not yet named, which is why Google is building AI models to map species ranges at scale (2025).
Source: Google DeepMind, Nov 2025
Exact wording at the source
With more than 2 million known species, and millions more to be discovered and named, that’s a monumental task.
Chart
Average decline in monitored wildlife populations by habitat type, 1970 to 2020
WWF Living Planet Index 2024, vertebrate populations.
- Freshwater85%
- Terrestrial69%
- Marine56%
Source: World Wildlife Fund, Catastrophic 73% decline in the average size of global wildlife populations in just 50 years reveals a 'system in peril'
Red List status and protected area coverage
- 948,646
The IUCN Red List included 172,620 species in October 2025, of which 48,646 were threatened with extinction.
Source: IUCN, Oct 2025
Exact wording at the source
The IUCN Red List now includes 172,620 species of which 48,646 are threatened with extinction.
- 1061%
61% of bird species worldwide have declining populations, up from 44% in 2016, according to the IUCN Red List update of October 2025.
Source: IUCN, Oct 2025
Exact wording at the source
Overall, 61% of bird species have declining populations – an estimate that has increased from 44% in 2016.
- 1111.5%
Of 11,185 bird species assessed by BirdLife International for the 2025 Red List, 1,256 (11.5%) are globally threatened.
Source: IUCN, Oct 2025
Exact wording at the source
Involving thousands of experts over nine years, 1,256 (11.5%) of the 11,185 species assessed are globally threatened.
- 1228%
The global green turtle population has grown approximately 28% since the 1970s, moving the species from Endangered to Least Concern in 2025.
Source: IUCN, Oct 2025
Exact wording at the source
Found in tropical and subtropical waters worldwide, the global population of green turtles has increased approximately 28% since the 1970s, despite ongoing threats to some subpopulations.
- 131,300
The IUCN Red List covered 1,300 fungi species in March 2025, of which at least 411 were at risk of extinction.
Source: IUCN, Mar 2025
Exact wording at the source
The addition of 482 newly assessed fungi species brings their number on the IUCN Red List to 1,300, of which at least 411 are at risk of extinction.
- 1417.6%
17.6% of land and inland waters and 8.4% of ocean and coastal areas were within documented protected and conserved areas in 2024, per the Protected Planet Report.
Source: UNEP-WCMC and IUCN, Oct 2024
Exact wording at the source
The Protected Planet Report 2024 reveals that 17.6% of land and inland waters and 8.4% of the ocean and coastal areas globally are within documented protected and conserved areas.
- 1521.6%
Reaching the 30% target by 2030 requires a further 12.4% of land and 21.6% of the ocean to be protected, according to UNEP-WCMC and IUCN (2024).
Source: UNEP-WCMC and IUCN, Oct 2024
Exact wording at the source
Over the next six years the global network will need to be urgently expanded by a further 12.4% on land and 21.6% in the ocean.
Chart
Share of bird species with declining populations
IUCN Red List and BirdLife International assessments.
- 201644%
- 202561%
Source: IUCN, Arctic seals threatened by climate change, birds decline globally: IUCN Red List
Camera traps and computer vision
- 161,295
The Wildlife Insights AI model is trained on over 35 million labeled camera trap images covering 1,295 species (2026).
Source: Wildlife Insights, 2026
Exact wording at the source
The Wildlife Insights AI model is trained with the largest curated dataset of labeled camera trap images to date, with over 35 million images across 1,295 species.
- 1780%
Blank images, often triggered by moving grass, can make up to 80% of a camera trap dataset, according to Wildlife Insights.
Source: Wildlife Insights, 2026
Exact wording at the source
Blank images can be up to 80% of a dataset and can take an expert several hours to sort through.
- 1894%
Wildlife Insights models identify 94% of blank camera trap images with an error rate of less than 2%.
Source: Wildlife Insights, 2026
Exact wording at the source
Filtering out blank images - the models are able to identify 94% of blank images with an error rate of less than 2%.
- 1997%
When an animal is in the image, Wildlife Insights models have a 97% chance of predicting an animal and a 91% chance of being correct at some taxonomic level.
Source: Wildlife Insights, 2026
Exact wording at the source
Overall, if there is an animal in the image, there is a 97% chance that the model will indeed predict some animal, and a 91% chance that the prediction will be correct in one of the taxonomic levels (including the “Animal” kingdom).
- 2065M
Google's open source SpeciesNet camera trap classifier was trained on a geographically diverse dataset of over 65 million images (released 2025).
Source: Google, Mar 2025
Exact wording at the source
SpeciesNet has been trained on a geographically diverse dataset of over 65M images
- 212000
SpeciesNet classifies images into more than 2,000 labels, including species, higher taxa and non-animal classes such as blank or vehicle.
Source: Google, Mar 2025
Exact wording at the source
It is designed to classify images into one of more than 2000 labels, covering diverse animal species, higher-level taxa (like "mammalia" or "felidae"), and non-animal classes ("blank", "vehicle").
- 2293.8%
Deep neural networks trained on the 3.2 million image Snapshot Serengeti dataset identified animals with over 93.8% accuracy across 48 species, in a study published in PNAS in 2018.
Source: Norouzzadeh et al., PNAS, Jun 2018
Exact wording at the source
Our deep neural networks automatically identify animals with over 93.8% accuracy, and we expect that number to improve rapidly in years to come.
- 2399.3%
By classifying only images it was confident about, the Snapshot Serengeti model automated 99.3% of identifications at the same 96.6% accuracy as volunteer teams.
Source: Norouzzadeh et al., PNAS, Jun 2018
Exact wording at the source
More importantly, if our system classifies only images it is confident about, our system can automate animal identification for 99.3% of the data while still performing at the same 96.6% accuracy as that of crowdsourced teams of human volunteers, saving more than 8.4 years (at 40 hours per week) of human labeling effort (i.e. over 17,000 hours) on this 3.2-million-image dataset.
- 2417,000 hours
Automating the Snapshot Serengeti labels saved more than 17,000 hours of human labeling, over 8.4 years of full-time work.
Source: Norouzzadeh et al., PNAS, Jun 2018
Exact wording at the source
More importantly, if our system classifies only images it is confident about, our system can automate animal identification for 99.3% of the data while still performing at the same 96.6% accuracy as that of crowdsourced teams of human volunteers, saving more than 8.4 years (at 40 hours per week) of human labeling effort (i.e. over 17,000 hours) on this 3.2-million-image dataset.
- 2596.2%
A 2025 study of 1.3 million images from 91 camera traps in two Spanish national parks reached an F1 score of 96.2% across 24 mammal species.
Source: Scientific Reports, May 2025
Exact wording at the source
We validated this strategy using 1.3 million images from 91 camera traps encompassing 24 mammal species and used 120,000 images for testing, achieving an F1-Score of 96.2% using expert models for final classification.
- 2690 percent
Wildlife Insights camera trap data gave more detailed range information than GBIF for 90% of mammals and nearly 50% of bird species with images in the platform, per a study reported by Conservation International in 2023.
Source: Conservation International, Jun 2023
Exact wording at the source
Wildlife Insights — a platform powered by an artificial intelligence (AI) model developed at Google and operated in partnership with Conservation International and others — provided more detailed information than GBIF on the geographic ranges of 90 percent of mammals and nearly 50 percent of the bird species with images in the platform, according to the study.
- 2711
11 rhino sites in Kenya use WWF and Teledyne FLIR thermal cameras with AI that detects people, wildlife or vehicles at night.
Source: World Wildlife Fund, 2025
Exact wording at the source
Currently, 11 rhino sites in Kenya are using FLIR cameras and equipment in some capacity, significantly reducing—and in some areas, completely halting—wildlife poaching.
Chart
Wildlife Insights AI performance on camera trap images
Wildlife Insights reported model accuracy, checked November 2026.
- Animal images flagged as animal97%
- Animal images correct at some taxonomic level91%
Source: Wildlife Insights, About Wildlife Insights AI
Bioacoustics and sound-based species monitoring
- 2850x
Perch helped the University of Hawaii's LOHE Lab find endangered honeycreeper sounds nearly 50 times faster than its usual methods (2025).
Source: Google DeepMind, Aug 2025
Exact wording at the source
Perch helped the LOHE Lab find honeycreeper sounds nearly 50x faster than their usual methods, enabling them to monitor more species of honeycreeper over greater areas.
- 29250,000
The first version of Google's Perch bioacoustics model was downloaded over 250,000 times between its 2023 launch and August 2025.
Source: Google DeepMind, Aug 2025
Exact wording at the source
Since it was first launched in 2023, the initial version of Perch has already been downloaded over 250,000 times and its openly available solutions are now well-integrated into tools for working biologists.
- 301.5 million
Perch 2.0, released in 2025, was trained on over 1.5 million labeled animal sound recordings.
Source: Google DeepMind, Aug 2025
Exact wording at the source
Our training set has over 1.5 million source recordings so we use a low-rank projection of rank 512.
- 3112 million
The Perch 2.0 embedding model has 12 million parameters, small enough to process large audio archives quickly.
Source: Google DeepMind, Aug 2025
Exact wording at the source
The embedding model is an EfficientNet-B3 ( Tan and Le, 2019 ) , a convolutional residual network with 12 million parameters, utilizing depthwise convolutions to maximize parameter efficiency.
- 3211k
BirdNET, built by the Cornell Lab of Ornithology and Chemnitz University of Technology, lists coverage of more than 11,000 species among its design priorities (2026).
Exact wording at the source
Noise robustness Real-time & Batch optimization High species coverage (11k+) Continuous data-driven updates
- 337,025
Rainforest Connection reports 7,025 distinct species identified and monitored with its acoustic tools, including 310 threatened species (2026).
Source: Rainforest Connection, 2026
Exact wording at the source
Distinct Species Identified & Monitored: 7,025 310 threatened species
- 34726,000
Rainforest Connection reports more than 726,000 hectares of rainforest protected with its acoustic monitoring (2026).
Source: Rainforest Connection, 2026
Exact wording at the source
Hectares of Rainforest Protected: 726,000+ 1 million football fields
Satellites and AI at sea and in forests
- 3572
72% to 76% of the world's industrial fishing vessels are not publicly tracked, according to a 2024 Nature study that used deep learning on satellite imagery from 2017 to 2021.
Source: Nature (Paolo et al.), Jan 2024
Exact wording at the source
We find that 72–76% of the world’s industrial fishing vessels are not publicly tracked, with much of that fishing taking place around South Asia, Southeast Asia and Africa.
- 3621
21% to 30% of transport and energy vessel activity is missing from public tracking systems, the same Nature study found.
Source: Nature (Paolo et al.), Jan 2024
Exact wording at the source
We also find that 21–30% of transport and energy vessel activity is missing from public tracking systems.
- 372 petabytes
The Nature study analyzed 2 petabytes of satellite imagery covering more than 15% of the ocean, where more than 75% of industrial activity is concentrated.
Source: Nature (Paolo et al.), Jan 2024
Exact wording at the source
To detect and characterize vessels and offshore infrastructure in coastal waters around the globe, we analysed 2 petabytes of satellite imagery spanning the years 2017–2021, with our analyses covering more than 15% of the ocean (Extended Data Fig. 1 ) in which more than 75% of industrial activity is concentrated ( Methods ).
- 3897%
The study's neural networks identified objects with over 97% accuracy and classified vessels as fishing or non-fishing with over 90% accuracy.
Source: Nature (Paolo et al.), Jan 2024
Exact wording at the source
We designed and trained three deep convolutional neural networks to identify objects (>97% accuracy) and estimate their lengths ( R 2 score of 0.84); to classify offshore infrastructure into oil, wind and other objects (>98% accuracy); and to classify vessels as fishing or non-fishing (>90% accuracy).
- 3953 billion
Researchers matched satellite detections against 53 billion vessel GPS positions to see which vessels were publicly tracked.
Source: Nature (Paolo et al.), Jan 2024
Exact wording at the source
We also analysed 53 billion vessel GPS positions from the AIS and matched them to the satellite detections to determine whether a detected vessel was publicly tracked.
- 4012
Global fishing fell by 12% at the start of the COVID-19 pandemic in 2020 and had not recovered by 2021, according to the satellite analysis.
Source: Nature (Paolo et al.), Jan 2024
Exact wording at the source
Globally, fishing decreased by 12 ± 1% at the onset of the COVID-19 pandemic in 2020 and had not recovered to pre-pandemic levels by 2021.
- 414.3 million
The world lost 4.3 million hectares of tropical primary forest in 2025, more than 11 football fields per minute, according to WRI's Global Forest Watch data.
Source: World Resources Institute, May 2026
Exact wording at the source
While the reduction is encouraging, the world still lost 4.3 million hectares of tropical primary forest in 2025, equivalent to more than 11 football (soccer) fields of forest per minute.
- 4236%
Tropical primary forest loss fell 36% from 2024 to 2025, after a record year of fires.
Source: World Resources Institute, May 2026
Exact wording at the source
Tropical primary forest loss decreased by 36% from 2024 to 2025 following a record-breaking year of extreme fires
- 4346%
Tropical primary forest loss in 2025 was still 46% higher than a decade earlier, per WRI.
Source: World Resources Institute, May 2026
Exact wording at the source
Tropical primary forest loss is still 46% higher than a decade ago.
- 4443%
Fires accounted for roughly 43% of global tree cover loss each year from 2023 to 2025, up from about 27% a year in 2001 to 2022, per WRI.
Source: World Resources Institute, Aug 2026
Exact wording at the source
Globally, fires accounted for roughly 43% of all tree cover loss each year between 2023 and 2025.
Chart
Fires as a share of global tree cover loss per year
World Resources Institute and University of Maryland data, published 2026.
- 2001 to 2022 average27%
- 2023 to 2025 average43%
Source: World Resources Institute, The latest data confirms: forest fires are getting worse
Citizen science and the conservation tech community
- 45100 million
eBird receives more than 100 million bird sightings each year from contributors around the world, according to the Cornell Lab of Ornithology (2026).
Source: Cornell Lab of Ornithology, 2026
Exact wording at the source
eBird is among the world’s largest biodiversity-related science projects, with more than 100 million bird sightings contributed annually by eBirders around the world and an average participation growth rate of approximately 20% year over year.
- 4620%
Participation in eBird grows at an average rate of approximately 20% a year.
Source: Cornell Lab of Ornithology, 2026
Exact wording at the source
eBird is among the world’s largest biodiversity-related science projects, with more than 100 million bird sightings contributed annually by eBirders around the world and an average participation growth rate of approximately 20% year over year.
- 47248
The first State of Conservation Technology report, led by WILDLABS and Colorado State University, surveyed 248 conservationists and technologists in 37 countries (2021).
Source: World Wildlife Fund, Dec 2021
Exact wording at the source
The report, led by WILDLABS and Colorado State University and supported by their NGO partners and tech sector leaders Microsoft and Arm, surveyed 248 conservationists and technologists across 37 countries using the WILDLABS.NET platform, asking them to rate 11 widely used tools for their capacity to advance conservation.
- 4890 percent
More than 90% of surveyed conservation technologists rated each of the top three emerging technologies, including machine learning and computer vision, as very helpful or game changers.
Source: World Wildlife Fund, Dec 2021
Exact wording at the source
More than 90 percent of respondents rated each of the top three emerging technologies as ‘very helpful’ or ‘game changers.’
- 4970 percent
GIS and remote sensing, drones and mobile apps had the highest performance ratings, with more than 70% of respondents rating them good or very good.
Source: World Wildlife Fund, Dec 2021
Exact wording at the source
The tools with the highest overall performance ratings were GIS and remote sensing, drones, and mobile apps, with more than 70 percent of respondents rating them ‘good’ or ‘very good.’
- 5050 percent
Wildlife Insights gave better range data than GBIF for nearly 50% of the bird species with images in the platform, according to the study reported by Conservation International (2023).
Source: Conservation International, Jun 2023
Exact wording at the source
Wildlife Insights — a platform powered by an artificial intelligence (AI) model developed at Google and operated in partnership with Conservation International and others — provided more detailed information than GBIF on the geographic ranges of 90 percent of mammals and nearly 50 percent of the bird species with images in the platform, according to the study.
FAQ
Questions readers ask
- How is AI used in wildlife conservation?
- AI mainly sorts and labels monitoring data: it identifies species in camera trap photos, recognizes calls in audio recordings and spots vessels or forest loss in satellite images. Wildlife Insights trains on over 35 million camera trap images, and a 2024 Nature study used deep learning to find that 72% to 76% of industrial fishing vessels are not publicly tracked.
- How accurate is AI at identifying animals in camera trap images?
- For well-sampled species, accuracy is high. Wildlife Insights reports a 97% chance of flagging an animal when one is present and a 91% chance of being correct at some taxonomic level, and the 2018 Snapshot Serengeti study reported over 93.8% accuracy. Accuracy is lower for rare species and regions with little training data.
- How many species does bioacoustic AI recognize?
- BirdNET lists coverage of more than 11,000 species, and Google DeepMind's Perch 2.0 was trained on over 1.5 million labeled recordings across birds, mammals, amphibians and other sounds. Rainforest Connection reports 7,025 species identified and monitored through its acoustic network.
- How much has wildlife declined?
- WWF's Living Planet Index shows an average 73% decline in monitored vertebrate populations between 1970 and 2020, with freshwater populations down 85%. The IUCN Red List counted 48,646 threatened species out of 172,620 assessed in October 2025.
- Does AI replace field biologists?
- No. AI handles the first pass on large datasets, such as removing blank images or flagging calls, and experts review the results. In the Snapshot Serengeti study, automating 99.3% of identifications saved over 17,000 hours of labeling, but confident predictions were still checked against human accuracy.
Sources
We used primary sources only: WWF and the Zoological Society of London's Living Planet Index, the IUCN Red List and BirdLife International, IPBES, UNEP-WCMC, Wildlife Insights and Conservation International, Google and Google DeepMind, the Cornell Lab of Ornithology, Rainforest Connection, the World Resources Institute, and peer-reviewed studies in Nature, PNAS and Scientific Reports. Population trend figures from the Living Planet Index are averages of relative change, not counts of animals. Red List totals grow as more species are assessed. Model accuracy figures come from each developer or study team, use their own test sets and are not directly comparable across tools. Counts reported by tool makers, such as species covered or hectares protected, are self-reported. Older baselines are labeled by year. Every quote was checked against the source page on the date shown. We store the sentence each number appears in, and a script reloads the publisher's page to confirm it is still there. The methodology has the details.
- BirdNET (Cornell Lab of Ornithology and Chemnitz University of Technology). About BirdNET (2026).
- Conservation International. Study: Camera traps key to closing biodiversity data gaps (Jun 2023).
- Cornell Lab of Ornithology. About eBird (2026).
- Google. SpeciesNet (cameratrapai) repository (Mar 2025).
- Google DeepMind. Mapping, modeling, and understanding nature with AI (Nov 2025).
- Google DeepMind. How AI is helping advance the science of bioacoustics to save endangered species (Aug 2025).
- Google DeepMind. Perch 2.0: The Bittern Lesson for Bioacoustics (arXiv) (Aug 2025).
- IPBES. Nature's Dangerous Decline 'Unprecedented'; Species Extinction Rates 'Accelerating' (May 2019).
- IUCN. Arctic seals threatened by climate change, birds decline globally: IUCN Red List (Oct 2025).
- IUCN. First 1,000 fungi on IUCN Red List reveal growing threats (Mar 2025).
- Nature (Paolo et al.). Satellite mapping reveals extensive industrial activity at sea (Jan 2024).
- Norouzzadeh et al., PNAS. Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning (Jun 2018).
- Rainforest Connection. Rainforest Connection: using sound and AI to protect rainforests (2026).
- Scientific Reports. Addressing significant challenges for animal detection in camera trap images: a novel deep learning-based approach (May 2025).
- UNEP-WCMC and IUCN. World must act faster to protect 30% of the planet: Protected Planet Report 2024 (Oct 2024).
- Wildlife Insights. About Wildlife Insights AI (2026).
- World Resources Institute. Latest analysis of deforestation trends (Global Forest Review) (May 2026).
- World Resources Institute. The latest data confirms: forest fires are getting worse (Aug 2026).
- World Wildlife Fund. Catastrophic 73% decline in the average size of global wildlife populations in just 50 years reveals a 'system in peril' (Oct 2024).
- World Wildlife Fund. Artificial intelligence and conservation (2025).
- World Wildlife Fund. First-ever State of Conservation Technology report identifies top 3 emerging technologies to advance conservation (Dec 2021).
Cite this page
Link to this page or to a single numbered figure. If you reuse a number, name the organization that first published it.
FruitPunch. (2026, November 15). AI for Conservation Statistics 2026. https://fruitpunch.ai/ai-for-conservation-statistics
Found a number that changed or a source that moved? Email editorial@fruitpunch.ai. We plan to check this page again by January 14, 2027.
Keep reading