The world has a plan to save insects - but almost no way to tell if it is working

Somewhere in a field in southern England, a camera sits on a tripod, photographing every moth that lands on a white sheet overnight. By morning, an AI system has identified each species from the images - no human expert required. Across the continent, weather radar designed to track storms is being repurposed to detect swarms of insects moving through the atmosphere. In backyards around the world, amateur naturalists photograph beetles and butterflies, uploading records that AI tools verify and scientists analyze.

These are the tools that might finally answer a question the world's governments committed to addressing but cannot yet measure: are insects recovering?

A plan with no feedback loop

In 2022, governments worldwide agreed to the Kunming-Montreal Global Biodiversity Framework - a set of 23 targets to protect and restore nature by 2030. The targets cover habitat protection, pollution reduction, pesticide management, and other measures that should, in theory, benefit insects. A global team of scientists examined those targets in a study published March 19, 2026 in Conservation Letters and concluded that the plan is well-designed. If implemented, it could help reverse falling insect numbers.

The catch is brutal: we have almost no way to tell if it is working.

The indicators governments use to track progress toward the 2030 targets are largely blind to insects. Dragonflies and damselflies are the only insect group to have been fully assessed for extinction risk globally. Out of more than one million named insect species - with an estimated four million more awaiting discovery - comprehensive data exists for one order. Everything else is, at best, patchwork.

"We could be doing all the right things and have absolutely no way of knowing whether it is working," said Lynn Dicks, a professor at the University of Cambridge and the study's senior author.

Falling at 1% per year, counted by almost no one

Previous studies have estimated that insect populations are declining by roughly 1% per year on average. That figure is alarming enough on its own, but its imprecision makes it even more troubling. We do not know which species are declining fastest, where the losses are concentrated, or whether any recovery is occurring in areas where conservation measures have been implemented.

The reasons for this data gap are practical, not philosophical. Insects are small, numerous, diverse, and often difficult to identify. Traditional monitoring requires trained entomologists to collect specimens, identify them under microscopes, and catalog the results - a labor-intensive process that cannot scale to the millions of species involved. Most countries do not have enough entomologists to monitor even their most common insects systematically.

Andrew Bladon, the study's lead author from the University of Reading, framed the technological shift as a solution to this fundamental scaling problem. "Technology is opening up possibilities that simply did not exist a decade ago. We can now monitor insects at a scale and speed that was unimaginable to previous generations of scientists."

AI cameras that never sleep

Automated camera traps represent perhaps the most immediately deployable tool. Set up in fields and woodlands, these systems photograph every insect that visits a light trap or bait station overnight. AI software processes the images, identifying species without human involvement. The approach generates far more data than a human surveyor could collect and operates continuously, night after night, across seasons.

The technology is not perfect. AI identification systems are only as good as the image libraries they were trained on, and those libraries are heavily biased toward well-studied groups in well-studied regions. A camera trap in a European woodland, where moth taxonomy is well-documented, will produce more reliable identifications than one in a tropical forest, where many species have never been photographed. But even imperfect data at scale is better than no data at all - and the training libraries improve with each year of deployment.

Radar, citizen science, and historical baselines

Weather radar offers a complementary approach that operates at a completely different scale. Radar systems already cover most of the world's temperate and tropical regions, and they routinely detect biological targets - birds, bats, and large insect swarms - alongside weather phenomena. Scientists are developing methods to extract insect movement data from existing radar archives, creating continental-scale pictures of aerial insect biomass over time.

The resolution is coarse. Radar cannot identify species. But it can detect large-scale trends - whether the total biomass of flying insects over a region is increasing, decreasing, or stable. For a field that currently lacks even this basic information for most of the world, radar-derived trends would represent a significant advance.

Citizen science contributes yet another data stream. Platforms where members of the public record insect observations have grown enormously in recent years, generating millions of records annually. When verified by expert review or AI tools, these records provide information on species distributions and phenology - the timing of seasonal events like emergence and migration - that no professional monitoring network could match in geographic scope.

Finally, the study points to existing research datasets collected by scientists and monitoring schemes over decades. These records, gathered independently by different teams for different purposes, could be pooled to create a historical baseline against which governments measure progress. The challenge is standardization: different surveys used different methods, different spatial scales, and different taxonomic focus, making direct comparison difficult.

What governments need to do

The study recommends that the United Nations establish a dedicated working group to develop insect-focused indicators for the biodiversity framework. This is not a small bureaucratic suggestion. Without specific, measurable indicators for insect populations, the 2030 targets will be evaluated primarily on the basis of vertebrate data - birds, mammals, fish - which may not reflect what is happening to insects at all. An ecosystem where bird numbers are stable but pollinator populations have collapsed is not a healthy ecosystem, even if the indicators say it is.

Bladon put the political dimension directly: "The big question is whether governments are willing to use new technology to hold themselves to account. A plan to save nature that cannot measure whether nature is actually recovering is not good enough."

What we still do not know

The study makes a strong case for what monitoring tools are available and how they could be deployed. It is less specific about the costs, institutional arrangements, and political will required to make deployment happen. Automated camera systems need to be purchased, installed, maintained, and their data managed - all of which requires funding that most national biodiversity programs do not currently have.

AI identification systems, while improving rapidly, still carry significant error rates for many insect groups, particularly in tropical regions where diversity is highest and reference libraries are thinnest. Building those libraries requires exactly the kind of expert taxonomic work that the automation is meant to supplement - a bootstrapping problem that will take years to resolve.

The 1% annual decline figure itself is an average across a limited number of studies, mostly from Europe and North America. Whether insect populations in Africa, Asia, and South America are declining at similar rates, faster rates, or different patterns entirely is largely unknown. The monitoring tools described in the paper could help answer that question, but only if they are deployed globally - a logistical and financial challenge of a different order from deploying them in a few well-resourced countries.

The 2030 deadline is less than four years away. If governments are serious about meeting their biodiversity commitments for the most diverse group of animals on Earth, the time to build the monitoring infrastructure is now.