American poultry farms are about to get a bird flu test they can run in the barn, with an answer in 45 minutes. Whether it amounts to a surveillance system depends on something harder to engineer: farms willing to share what they find.

A grower swabs a bird’s throat, drops the swab into a buffer vial, draws 400 microliters up a plastic pipette and dispenses it into a cartridge. The cartridge goes into a battery-powered reader that runs four hours on a charge. Forty-five minutes later the reader reports whether the bird is infected with influenza A and, if it is, whether the virus carries H5 or H7, the subtypes that can turn highly pathogenic.

Alveo Technologies expects USDA approval for that device in the coming months. Fall migration is already moving waterfowl down the flyways with outbreaks reported from commercial flocks in Minnesota, Wisconsin and both Dakotas . Since February 2022 the virus has reached 1,985 American poultry premises and cost more than 185 million birds, by USDA’s count .

The New Part Is The Engineering, Not The Chemistry

The assay runs on reverse transcription loop-mediated isothermal amplification, a process which copies viral genetic material at a single temperature and skips both the RNA extraction and the thermal cycling required for PCR, the process used in most conventional diagnostics. Alveo published a laboratory validation with its Dutch partner Royal GD. Daniel Perez, who is the Caswell S. Eidson Chair in Poultry Medicine and Georgia Research Alliance Distinguished Investigator at the University of Georgia, ran independent challenge trials on the assay. The novelty, as he sees it, is in the electronics. LAMP chemistry dates to 2000 , and many groups have built avian influenza assays on it. Alveo’s advantage, Perez told me, is that “it ditches optical detection entirely in favor of semiconductor-based sensing.”

Optics are the expensive, breakable part of a molecular test. Reading the reaction electrically instead lets Alveo put seven independent assays on one cartridge: two for influenza A, three for H5 lineages including clade 2.3.4.4b, one for H7 and one internal control. One swab yields seven answers, on an analyzer that costs under $500.

A Narrower Detection Window, On Purpose

That validation reports 100 percent specificity against 12 influenza A strains and 19 non-influenza organisms, with a limit of detection equivalent to PCR cycle thresholds of 29 to 33. A benchtop RT-PCR machine reaches further, out past Ct 38, and Perez argues that the extra reach buys a farm very little. Those faint signals, he said, “often fall outside the window of peak transmissibility or indicate residual viral fragments from a cleared infection.”

A test that stops at Ct 33 finds the birds shedding enough virus to infect the barn. It misses birds that have already cleared, which matters more to an epidemiologist reconstructing a chain than to a grower deciding what to do this afternoon.

Alveo announced a partnership last month with AgriNerds , a start-up co-founded in 2018 and run by UC Davis poultry epidemiologist Maurice Pitesky, whose Waterfowl Alert Network combines empirical HPAI detections with weather radar, satellite imagery and telemetry to predict daily waterfowl abundance around roughly 160,000 commercial poultry farms at 250 meter resolution. Nearly all of it runs on taxpayer-funded infrastructure: NEXRAD weather radar, government satellites and USDA-supported research. When the model forecasts a spike near a farm, the grower’s app tells him to test.

USDA has classified 1,947 detections by phylogeny and attributes 1,329 of them to wildlife introductions, though farm-to-farm spread passed them in 2025 due to large clusters in Indiana and Ohio. Daily testing would spread cartridges evenly over a calendar of mostly safe days. “You don’t want to just run our tests blindly like a peanut butter spread running a test every day of the year,” Erik Tyrrell-Knott, Alveo’s president and chief business officer, told me.

When to test also depends on what the farm raises. Perez ran clade 2.3.4.4b in mallards and in chickens, and found the two species call for opposite strategies. Ducks, he said, “are highly susceptible to infection but resistant to clinical disease,” which makes them “silent amplifiers of the virus.” Waiting for sick birds loses the barn: “Relying on mortality or symptom-driven testing in duck operations will fail.” This suggests duck operations should swab healthy birds on a fixed schedule, whatever the radar says.

Chickens, most of American poultry, show the reverse . In his trials they resisted infection at higher doses but crashed quickly once infected, with sudden mortality and rapid egg drops. For them Perez says “continuous asymptomatic swabbing is an inefficient use of resources,” implying that tests should be reserved until clinical anomalies are observed. In fact, the design problem for optimal surveillance of influenza in poultry has not yet been solved.

That leaves the waterfowl alert as a possible third trigger for testing, somewhere between a standing schedule and waiting for a dead bird. But the positive predictive value of waterfowl testing is low, Pitesky told me: “we see these spikes very often in waterfowl abundance around farms, but we don’t see an outbreak.” Growers subscribe anyway, because a high-risk day gives them a reason to restrict who enters the barns or move a processing date.

I think they are right to. An alert does not have to perfectly forecast an outbreak to be worth having. In the case of testing, it simply has to tip a cheap decision at a moment when being wrong costs little. Restricting access is cheap. Depopulating a layer house is not.

More could be done. Sharpening the alert would take farm-level outbreak data, but the industry will not release it. This is a collective action problem. About 1,500 farms in 25 states have opted into Pitesky’s network, compared with the 160,000 farms his model covers. USDA and the state agriculture departments, he says, have protected the poultry industry’s data “to an illogical conclusion,” leaving only county-level numbers in public, too coarse to relate individual farm outbreaks to the waterfowl activity around them. His suggestion: withhold indemnification from farms that decline to share data with researchers under agreements that protect its privacy.

The chemistry is expected to clear USDA in the coming months. The data will take longer, because no farm would gain much by being the first to open its records, although every farm gains if they all do. Two things could break the deadlock. The industry can build a network whose warnings are worth more than the secrecy they cost, or USDA can attach conditions to the indemnity checks it is already writing.