An insulation crew finishes a 2,400-square-foot house in a day, maybe two. They cut batts, stuff cavities, staple vapor barriers, move on. Nobody looks back. The drywall crew shows up 48 hours later and covers everything, and from that moment forward, the only way to know whether the insulation was actually installed correctly is to either tear open the wall or wait for the first winter heating bill to arrive 40 percent higher than it should be.
There is a third option that costs $250 and takes less time than a lunch break.
What the Camera Sees That You Cannot
A thermal imaging camera does not see insulation. It sees surface temperature differences, which, when the building is pressurized or when there is even a modest temperature differential between inside and outside, reveal exactly where insulation is missing, compressed, or installed with gaps that defeat its purpose entirely. A properly insulated stud bay shows as a uniform color field. A bay with a void shows as a sharp thermal gradient, a cold stripe cutting across what should be an even surface, as obvious in infrared as a missing tooth is in a smile.
FLIR's ONE plugs into the USB-C port on your phone. It costs $199 to $399 depending on the model, weighs a few ounces, and its native app walks you through a room-by-room scan with step-by-step guides that assume you have never held a thermal camera before, because most people have not. Point it at a wall. Sweep slowly. Its software overlays infrared and visible-light images so you can see both what the wall looks like and where the heat is going.
Professional standalone units from Hikmicro, FLIR, and InfiRay range from $500 to $1,000 and offer higher resolution, which matters when you need to distinguish between a compression defect and a complete void, or when the temperature differential is small enough that a lower-resolution sensor misses it. Pistol-grip models above $1,000 measure temperature differences under half a degree Celsius, sufficient to find thermal bridges through framing members that basic models treat as noise.
Half of New Homes Fail, and Nobody Checks
A foundational NIST field study that examined insulation performance across newly weatherized homes found results bad enough to question whether quality control in residential insulation installation exists at all. Roughly 30 percent of homes had insulation missing from 5 to 10 percent of their wall area, which sounds minor until you calculate that even on a modest 1,600-square-foot house, 5 percent is 80 square feet of bare wall bleeding heat. About half the homes had smaller gaps, under 5 percent of wall area, but those still violated DOE installation standards. Shrinkage and fissures appeared in two-thirds of inspected homes. Heat losses at wall-to-wall joints showed up in 74 percent. Window heat losses in 72 percent.
Add those numbers up and over two-thirds of all thermographically inspected dwellings had serious defects after professional installation.
RESNET, the Residential Energy Services Network, grades insulation installations on a three-tier scale. Grade I means minor defects only, gaps under 2 percent of area, no voids extending through the full cavity width. Grade I is the minimum for ENERGY STAR certification. Grade III means substantial defects: missing sections, severe compression, gaps that essentially negate the insulation's purpose. The Insulation Institute acknowledges that while Grade I "should be the norm in the residential building industry, there are instances where it is not," a polite way of saying the failure rate is high enough that the industry's own standards body felt compelled to publish remedial guides.
Most new homes never receive a thermal scan. The building inspector walks through with a flashlight, confirms batts are present in cavities that are visible, and signs off. What the flashlight cannot tell you: whether the batt was cut to fit around the junction box or stuffed and compressed against it, whether the vapor barrier was sealed or just stapled across the face of the stud with gaps at every penetration, whether the rim joist cavities in the basement got insulation at all or whether the crew forgot them entirely because they are hard to reach and nobody ever looks.
A thermal camera answers all of those questions in the time it takes to walk the house once.
Where AI Changes the Economics
Knowing how to operate a thermal camera is trivial. Knowing how to interpret what it shows you is the expensive part. A certified thermographer bills $150 per hour or more, understands emissivity corrections, knows that a south-facing wall heated by afternoon sun will produce thermal gradients that look like insulation defects but are not, can distinguish a moisture intrusion pattern from a missing batt pattern based on the shape and distribution of the anomaly. This expertise is real and expensive and, until recently, was the only way to get reliable results from a thermal scan.
That barrier is falling. Researchers at Umeå University in Sweden published a method using generative adversarial networks to predict what a building's thermal distribution should look like based on its visible-light appearance, then flagging any region where the actual thermal image diverges from the prediction as an anomaly. No labeled training data of defects required, no thermographer needed to mark up examples. It learns what "normal" looks like and catches everything that is not, which is precisely the pattern-matching task at which machine learning excels and which human inspectors, scanning dozens of images after a long day, occasionally miss.
Separately, a deep learning framework for panoramic infrared inspection demonstrated fully automated analysis of building facades in roughly ten minutes, including semantic segmentation that identifies window frames, wall sections, and structural elements, then quantifies the percentage of each surface area exhibiting thermal anomalies above a calibrated threshold. A human inspector doing the same analysis on the same dataset takes hours and produces results that vary depending on their experience, their fatigue, and whether they had coffee.
FLIR's own trajectory tells the commercial story. Their Ignite cloud platform already stores and organizes thermal images; the FLIR Home Inspect app provides guided workflows for untrained users. Market analysis from Teledyne FLIR projects that AI-augmented thermal products will account for more than 40 percent of new product launches by 2027, with subscription-based analytics services that process images in the cloud and return annotated reports. You shoot, the cloud interprets, the defects get flagged with confidence scores and remediation recommendations.
GSA tested drone-mounted thermal cameras on federal buildings and found they detected 50 percent more defects than manual evaluations while costing 80 percent less and finishing 80 percent faster. Residential drones are smaller, cheaper, and increasingly autonomous; a thermal scan of every exterior wall of a two-story house from the air takes a single flight of under 15 minutes.
The ROI Math for a Builder
A pre-drywall thermal scan costs $280 on average nationally when you hire an inspector, according to Angi's 2026 data. If you own a FLIR ONE ($250 one-time) and spend 20 minutes scanning a house yourself, the marginal cost per inspection after the first house is effectively zero. Even at the hired-inspector rate, the math is brutal in your favor.
Consider what happens when insulation defects go undetected. Fixing insulation in a finished wall costs $1,150 to $6,000 per wall section, because you are paying to remove drywall, correct the insulation, then replace and refinish the drywall. Catching it before drywall costs the insulation crew ten minutes of repositioning batts. That is the difference between a $3,000 callback and a $0 fix.
Construction rework consumes 5 to 10 percent of total project costs industrywide, an estimated $65 billion annually according to an Autodesk and FMI joint analysis, with the Navigant Construction Forum estimating 9 percent when indirect costs like supervision and schedule delays are included. Insulation defects are among the most common and most preventable categories of rework, because the failure mode is nearly always the same: gaps, voids, compression. Visible in infrared. Invisible once the drywall goes up.
For a production builder doing 50 homes a year, a $250 camera that prevents even two warranty callbacks per year has paid for itself before the end of the first quarter.
The Strongest Case Against
Thermal imaging has a physics problem. Pre-drywall scans happen when the building envelope is open, which often means minimal temperature differential between inside and outside, which means the camera has less signal to work with. ASTM C1060, the standard governing thermographic insulation inspection, explicitly requires a temperature difference for reliable results, historically at least 10°C (18°F) between interior and exterior air. On a 75°F day with no HVAC running in an open-framed house, you are not going to see much.
Workarounds exist. Blower-door depressurization pulls outside air through every gap in the envelope, creating detectable temperature differentials even in mild weather. Some inspectors run a portable heater for several hours before scanning. But these add time and complexity to what is supposed to be a quick, cheap quality check, and they are exactly the kind of procedural steps that get skipped on a busy production homesite where the drywall crew is already parked outside waiting.
AI interpretation is also younger than its marketing suggests. Umeå University's GAN-based anomaly detector was demonstrated on commercial facades, not residential stud-wall assemblies, and the ten-minute automated analysis framework was calibrated on a 1996 reinforced-concrete building in South Korea, not a 2x6 wood-frame house in Phoenix. Translating academic results into a tool that works reliably across the enormous variety of American residential construction, from SIP panels in Minnesota to ICF walls in Florida to standard 2x4 frames in everywhere else, is a significant engineering challenge that nobody has yet shipped as a consumer product.
And a thermal camera cannot tell you why a spot is cold. It shows that something is wrong; it does not show what. A dark spot on the scan could be a missing batt, a compressed batt, a wet batt, an air leak through an unsealed penetration, or a thermal bridge through a steel bracket. Diagnosing the root cause still requires someone who understands building science, which means the AI can reduce the expertise needed but cannot yet eliminate it.
What to Do With This
If you are buying new construction, ask your builder whether they perform thermal imaging before drywall. If the answer is no, hire a pre-drywall inspector who carries a thermal camera. Budget $280. It is the single best insurance policy available at that stage of construction, cheaper than a single day's interest on your mortgage and capable of catching defects that will otherwise cost thousands to fix and will certainly raise your energy bills for every year you own the home.
If you are a builder or site superintendent, buy a FLIR ONE for $250 and run it across every house after insulation and before drywall, every single time. You do not need to be a thermographer. You need to see obvious voids. That bright-blue stripe cutting across an otherwise uniform yellow wall is not subtle. It is screaming at you. Point it out to the insulation crew, have them fix it while the walls are still open, and move on. Total time cost: 15 to 25 minutes per house. Total savings: every insulation callback you will never receive.
If you are an insulation contractor, consider making thermal verification part of your standard delivery. A crew that hands the builder a thermal scan showing clean, uniform coverage alongside the insulation certificate is a crew that wins the next bid. The competitive advantage of verifiable quality in a trade where quality is almost never verified is enormous, and the cost of providing it is a $250 camera and 20 minutes per job.
Watch the AI analysis tools. FLIR's cloud analytics pipeline, drone-mounted thermal scanning, and emerging smartphone apps that flag anomalies automatically are all converging on the same destination: a world where every house gets scanned, every defect gets caught, and the question of whether your insulation was installed correctly stops being a mystery you solve with your first heating bill and starts being a fact you verify before the drywall truck arrives.
What This Analysis Does Not Cover
The NIST defect rates cited above come from weatherized homes, not exclusively new construction, and the study is old enough that construction practices, materials, and standards have changed substantially since publication. No comparable large-sample thermographic study of newly built homes under current energy codes exists in the public literature, which is itself a finding: the industry has not measured the problem at scale, likely because doing so would produce numbers that nobody in the insulation business wants to publish.
The ROI calculation assumes a defect detection rate based on the NIST data, which may overstate the problem for modern construction with current RESNET inspection requirements. Homes built to ENERGY STAR specifications with Grade I verification by a HERS rater have already passed a form of quality control that most homes never receive, making the marginal value of a thermal scan on those homes lower.
AI-powered interpretation of residential thermal scans is not yet a commercially available product in the way this article suggests it soon will be. Academic research is promising; the commercial reality is that FLIR's app provides guided scanning, not automated defect diagnosis. That 40-percent-of-new-launches forecast describes the broader thermal imaging market, not residential construction specifically. The gap between what researchers have demonstrated in papers and what a site superintendent can download from the App Store remains substantial.
And none of this replaces a blower-door test, which measures the actual air leakage rate of the entire building envelope quantitatively, not just where it leaks visually. Thermal imaging and blower-door testing are complementary tools. The camera shows you where; the door tells you how much.