Industries / Construction and property / Pest control
Inspection and identification
Finding where pests live in a building or on a property and naming the species correctly, because the right response depends on what the pest is.
5
Milestones achieved, from 1945 to 2026.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1945
A handbook for the trade
Arnold Mallis, an entomologist, published Handbook of Pest Control: The Behavior, Life History, and Control of Household Pests in 1945. As its subtitle says, it covered the behavior, life history and control of household pests in one book. Open Library catalogs the 1945 edition at 554 pages and lists new editions in 1954, 1960, 1964, 1969, 1982, 1990 and 1997; the 1997 eighth edition, edited by Dan Moreland, runs to 1,456 pages.
Credit
Arnold Mallis
2008
Dogs trained on bed bugs
Bed bugs hide in cracks and are hard to see, so entomologists at the University of Florida trained dogs to find them by smell. In their tests the dogs picked out bed bugs from carpenter ants, German cockroaches and termites with a 97.5% positive indication rate and no false alerts. They told live bed bugs and viable eggs from dead bugs, cast skins and feces 95% of the time, with a 3% false alert rate on feces, and found live bed bugs hidden in hotel rooms with 98% accuracy (Journal of Economic Entomology, 2008).
Credit
Margie Pfiester, Philip G. Koehler and Roberto M. Pereira, University of Florida
2010
Infestations with no insects
Some people are certain they are infested by bugs that no one else can see. Nancy C. Hinkle, an entomologist at the University of Georgia, reviewed this condition, Ekbom syndrome, also called delusory parasitosis, in the Annual Review of Entomology in 2010. She wrote that sufferers seek identifications and treatment from physicians and entomologists, that the belief cannot be corrected by argument or evidence, and that treatment usually needs medical professionals. In 2011 she wrote that entomologists estimate more than 100,000 Americans have it.
LessonWhen no specimen can be found and named, the problem may not be a pest at all, and pest treatment will not end it.
Credit
Nancy C. Hinkle, University of Georgia
2014
Dogs falter in real apartments
Rutgers entomologists tested 11 dog teams in apartments with natural bed bug infestations. Every handler believed their dog found infestations at least 95% of the time. Across three experiments the dogs found 44% of infestations on average (from 10% to 100%, team by team), with an average false alert rate of 15% (from 0% to 57%). Teams tested on several days were inconsistent from day to day, and neither experience nor certification predicted how well a team did (Journal of Economic Entomology, 2014).
LessonAccuracy measured with planted bugs does not carry over to occupied homes until it is measured there.
Credit
Richard Cooper, Changlu Wang and Narinderpal Singh, Rutgers University
2026
Phone apps misname ticks
CDC scientists tested three free phone apps with built-in AI identification, Google Lens, Seek by iNaturalist and Obsidentify, on five species of ticks that bite people in the United States, using five phones and five users. Across species, life stages and feeding states, the apps identified 54% of specimens as ticks and 16% to the correct species. They did worst on unfed young ticks and on ticks swollen with blood (Journal of Medical Entomology, September 2026).
LessonAn identification app has to be tested on the life stages and look-alikes people actually find before anyone relies on it.
Credit
James C. Burtis, Rebecca J. Eisen and colleagues, CDC Division of Vector-Borne Diseases
Watch for
Swab strips for bed bugs
A lateral flow strip, the technology of a home pregnancy test, detects a bed bug protein in residue swabbed from beds and furniture; it is sold under the brand TruDetx. In 27 apartments with light infestations (fewer than 20 bed bugs caught in pitfall traps over 14 days), the strip identified 96% of infested units, against 40% for two technicians' quick visual inspections and 32% for two dog teams. The strip cannot tell old residue from new: in the author's further tests, residue from 10 fed bed bugs took 90 days to fall below detectable levels.
Achieved whena team with no tie to the strip's developers or seller repeats the field comparison in other buildings and publishes its counts.
Claimed by
Alexander Ko of Envu, whose 2020 laboratory study with Dong-Hwan Choe of the University of California, Riverside developed the test, in the proceedings of the International Conference on Urban Pests (ICUP), June 2025
Bed bugs found in air
Researchers at CSTB, France's scientific and technical center for building, listed the volatile organic compounds (VOCs), chemicals that evaporate into the air, that bed bugs give off. They then sampled the air in social housing in and around Paris between October 2019 and April 2021. A decision tree, a simple rule-based model, was trained on 40 of these chemicals in 102 dwellings and tested on 41 others; the authors report it detected bed bugs correctly 93% of the time and predicted their absence correctly 95% of the time. The work led to patent application WO 2024/069385 A1.
Achieved whenanother laboratory repeats the air test in homes it did not choose, or a commercial service publishes results from routine use.
Claimed by
Isabelle Lacaze and colleagues, CSTB, in the ICUP 2025 proceedings, June 2025
Photo ID for technicians
NameDatBug launched a phone app for pest management professionals that returns a species identification, a confidence score and treatment guidance from a photo within seconds. It includes integrated pest management (IPM) protocols for more than 700 species. The announcement gives no accuracy figure.
Achieved whenan independent test on expert-confirmed specimens publishes the app's accuracy for each group of pests.
Claimed by
NameDatBug, developed by Adam Holt and Rob Fryer, July 2026
Open
01
Bed bug look-alike test set
A public set of phone photos of bed bugs at every life stage, with their eggs, cast skins and fecal spots, and of the insects most often mistaken for them, such as bat bugs, carpet beetles and their larvae, cockroach nymphs, booklice, spider beetles, fleas and ticks. An entomologist confirms each specimen under a microscope. A scoring sheet records what free identification apps and general AI models say about every photo and reports accuracy by group and life stage.
The briefCredit · unclaimed02
Name droppings from a photo
A test set and an open model that tell mouse, rat, cockroach, bat and lizard droppings apart in phone photos taken beside a scale card, and that say when they cannot tell. Score three pest technicians on the same photos as a baseline.
The briefCredit · unclaimed
Open 01 · the brief
Bed bug look-alike test set
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of phone photos of bed bugs at every life stage, with their eggs, cast skins and fecal spots, and of the insects most often mistaken for them, such as bat bugs, carpet beetles and their larvae, cockroach nymphs, booklice, spider beetles, fleas and ticks. An entomologist confirms each specimen under a microscope. A scoring sheet records what free identification apps and general AI models say about every photo and reports accuracy by group and life stage.
Why it is still open
CDC scientists ran this kind of test for ticks in 2026 and found that three apps named only 16% of specimens to species. Searches of Europe PMC, Crossref, GitHub, Hugging Face and Zenodo found no such test for bed bugs. The closest public set, BedBugScan on Roboflow (CC BY 4.0), holds 266 photos of bed bugs, eggs, molts and droppings and no look-alike insects. A 2025 set of more than 25,000 insects caught on light-trap sticky cards in food premises is labeled only to order, such as flies or beetles. NameDatBug, launched in July 2026, publishes no accuracy figure.
Built with what exists
Two or three phones with clip-on macro lenses; a stereo microscope; preserved specimens borrowed from a university insect diagnostic lab or a natural history museum, so no live insects are needed; Label Studio for labels; Google Lens and Seek by iNaturalist, run by hand on each photo; general AI models that read images, such as Claude or Gemini, called by a Python script.
First step
Ask a university extension diagnostic lab or a natural history museum for an afternoon with preserved bed bugs and five look-alikes, and photograph 50 specimens with two phones.
Done looks like
At least 1,000 photos covering bed bugs and at least eight look-alike groups, every identity confirmed under a microscope; for each tool, the share of bed bugs it missed and the share of other insects it called bed bugs, by life stage; all photos released under an open license.
Open 02 · the brief
Name droppings from a photo
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A test set and an open model that tell mouse, rat, cockroach, bat and lizard droppings apart in phone photos taken beside a scale card, and that say when they cannot tell. Score three pest technicians on the same photos as a baseline.
Why it is still open
Searches of Europe PMC, Crossref, GitHub and Hugging Face found no published classifier or accuracy test for pest droppings. The nearest work grades the consistency of laboratory rats' feces (Mori and colleagues, BMC Research Notes, 2026). Consumer apps such as Droppings Identifier (version 1.0, January 2025) name animal droppings from a photo, but give no accuracy figure and do not mention household pests.
Built with what exists
Phones and a printed scale card; photos taken in place by pest technicians at sites where a trap catch or a camera has confirmed the animal; Label Studio for labels; an open image model such as Ultralytics YOLO11; general AI models that read images, as a comparison.
First step
Agree with one pest control company that its technicians will photograph droppings in place, beside a scale card, at jobs where a trap or camera confirms the pest. Nobody should sweep up or handle the droppings: CDC warns that stirring up droppings of infected rodents can put hantavirus into the air.
Done looks like
1,500 photos from at least 50 sites, each tied to a confirmed species; accuracy for each kind of dropping, the share of photos the model declines to call, and the technicians' accuracy on the same photos, all published with the photos.