Matt Belcher builds high-performance homes in Wildwood, Missouri, certifying every project to the ICC 700 National Green Building Standard and installing geothermal systems, triple-pane windows, and spray foam insulation rated well beyond code, because his clients are paying for a house that performs like infrastructure rather than a house that performs like a code-minimum box with nicer finishes. Then the appraiser arrives, glances at the comparable sales in the MLS, and values the home as if none of those features exist.
So Belcher stopped losing this fight: he now requires, in every sales contract, that any lender's appraiser hold at least 14 hours of green valuation education from a qualified provider, preferably carrying an MAI or SRA designation from the Appraisal Institute. One recent project came back undervalued. He invoked the clause, demanded a second appraisal, and the number came back "reflecting a much fairer value," according to NAHB's case study.
Most builders do not write contract clauses to protect themselves from their buyer's appraiser. Most have no idea they need to. Belcher does, because he has watched the same invisible erasure happen on deal after deal after deal, and he got tired of it.
The Data the Algorithm Learns From
On May 5, 2026, ATTOM launched what it called a "ground-up, AI-first rebuild" of its automated valuation model, covering 98 million U.S. properties with a median absolute percentage error of 2.9 percent, meaning more than half the time the model's estimate lands within 2.9 percent of the price a human being actually paid at a closing table. Impressive. Aaron Wagner, ATTOM's VP of Data Science, described the system as one that "models how each neighborhood has evolved over the past 30 years" to translate historical sales into present-day values.
Thirty years, and that is precisely the problem.
For most of that span, geothermal heat pumps were exotic. Spray foam was a specialty product that most insulation contractors did not carry. Triple-pane windows were a European curiosity that American manufacturers dismissed as unnecessary for our milder climates, and HERS ratings existed but penetrated a fraction of new construction, and solar panels were rare enough to serve as conversation pieces rather than infrastructure. And the MLS, the transactional database that feeds every automated valuation model in the country, had no standardized fields for any of it, which means that for three decades the single largest source of residential pricing data in the United States was structurally incapable of recording whether a home's envelope leaked like a screen door or held pressure like a submarine.
Sandra Adomatis, perhaps the most cited expert on green home appraisal methodology, recently pulled 30 comparable sales from her local MLS and searched for green or energy-efficient identifiers. Found none. Not because the homes were all conventional builds, but because nobody entered the data, because the fields either did not exist, or the listing agent skipped them, or the sale closed before anyone thought to ask whether the $38,000 geothermal loop buried in the backyard might be worth documenting so that the next buyer's lender could see it and the next buyer's appraiser could price it and the next transaction in the MLS could finally, for once, record that ground-source heating is a thing that costs real money and saves real money.
"The current system structurally suppresses value," Adomatis told Green Builder Media. "If the market can't see it, it can't price it. If appraisers can't find it, they can't adjust for it. If lenders can't verify it, they can't underwrite it confidently."
She kept using it.
How 2.9% Accuracy Becomes a $42,000 Blind Spot
Consider a home that costs $580,000 to build. Say the builder installs a ground-source heat pump ($18,000), spray foam insulation throughout ($8,500), ERV ventilation with CO2 sensors ($4,200), triple-pane low-E windows ($6,800), and a 9.6 kW solar array with battery backup ($4,500 net after federal tax credit), bringing the total green premium to roughly $42,000.
ATTOM's AVM analyzes comparable sales in the neighborhood, finds conventional builds that sold for $530,000 to $555,000 over the past two years, sees no reason, trained as it is on 30 years of transactions that never recorded green features, to adjust upward, and spits out a valuation of $548,000. A 2.9 percent median error on $548,000 is about $16,000 in either direction.
Forty-two thousand dollars in green upgrades. The model's entire confidence interval cannot reach them. Not close.
This is not a bug. The model is doing exactly what it was designed to do: reflect what the market has historically paid for homes with similar recorded characteristics, which sounds reasonable until you realize that the characteristics distinguishing a high-performance home from a code-minimum home were never in the record to begin with, because the database that every AVM in America relies on was not designed to capture them and nobody with the authority to fix that omission has bothered to. Accurate? Sure. In the way a thermometer is accurate when you hold it in the shade and ask what the temperature is in the sun.
The Federal Rule That Might Not Help
On July 17, 2024, six federal agencies published a final rule requiring quality control standards for AVMs used by mortgage originators and secondary market issuers. Mandated by Section 1125 of Dodd-Frank and more than a decade overdue, the rule imposes five requirements: ensure high confidence in estimates, protect against data manipulation, avoid conflicts of interest, require random sample testing, and comply with applicable nondiscrimination laws.
That last factor is an independent requirement, meaning institutions must build specific policies to ensure their AVMs do not violate fair lending laws. It became effective in the fourth quarter of 2025.
On paper, this should pressure AVM providers to examine their models for systematic bias. In practice, the rule says nothing about green features, does not require AVMs to recognize energy-efficient construction, and does not define what "high confidence" means when the training data systematically excludes an entire category of home improvement. That fifth factor targets racial and demographic bias in property valuations, a documented and serious problem with decades of evidence behind it, but it does not extend to construction methodology bias, and a model that consistently undervalues ICF homes or net-zero homes or homes with standing-seam metal roofs rated to withstand 180 mph winds is not, under any current reading of the law, discriminating against a protected class.
One exception deserves attention, however: if green-built homes are concentrated in neighborhoods that also have racial or socioeconomic characteristics protected under fair lending law, then the AVM's inability to value green features could produce disparate impact. That intersection has not been tested, and it should be.
Congress Noticed. Then Moved On.
In May 2024, Representative Sean Casten of Illinois and Senator Michael Bennet of Colorado introduced the GREEN Appraisals Act, which would require residential appraisals to consider energy efficiency characteristics, renewable energy features, estimated energy savings, and energy consumption relative to comparable properties. It would also mandate additional appraiser education and training in green valuation.
It has not advanced. In a Congress that has not passed major housing legislation since the Dodd-Frank Act itself, a bill requiring appraisers to count solar panels is not setting the agenda. Builders have been waiting for legislative help with green appraisals for over a decade, and that decade has produced exactly one meaningful industry tool: the Appraisal Institute's Residential Green and Energy Efficient Addendum, a worksheet that lets builders catalogue performance certifications, ratings, and features across categories including indoor air quality, daylighting, and insulation values, developed years ago by an organization that understood the problem long before Congress noticed it. Bill Garber, the Institute's senior director, frames it simply: "Appraisers can't support what they can't document."
He is right. The tool exists and the training exists, but neither has caught on at scale, because catching on at scale would require either a regulatory mandate or a market incentive, and neither exists.
The Resilience Dimension
At the 2025 International Builders' Show, Cindy Wasser of Home Innovation Research Labs announced that NGBS Green certification now includes "Green+" resilience add-ons covering wildfire, wind, water, and earth-movement resilience. These are pursued as part of standard NGBS certification, not as separate processes, which reduces friction.
Sealed roof decks, ember-resistant vents, impact-rated glazing, structural tie-down systems that exceed code: these features earn Green+ designation, and every single one of them is invisible during a typical appraisal inspection, invisible because you cannot see a sealed roof deck from the driveway, because an ember-resistant vent looks identical to a standard one from six feet, because a structural continuous load path from foundation to ridge is hidden behind drywall before the appraiser walks through the front door.
In wildfire zones, these features decide whether a home survives or burns. In the appraisal, they are worth exactly as much as the code-minimum ranch down the street that would ignite from a single ember landing in its soffit vent.
Strongest Counterargument
AVM providers would argue, and they are not entirely wrong, that their models reflect market value, not replacement cost or feature value. Fair enough. If buyers in a given market do not pay a measurable premium for green features, then an AVM that invents a premium would be less accurate, not more, and the fix belongs in MLS data standards, in listing agent behavior, in buyer education, and in the slow accretion of enough documented green-home sales to create the statistical foundation these models need before they can learn something they have never been shown. An algorithm cannot learn from data that does not exist.
Correct, and also a perfect closed loop. Green features are not valued because they are not in the data. They are not in the data because they are not tracked, not tracked because nobody enforces tracking, not enforced because the market does not appear to value them. And the machine learns the loop, hardens it into mathematics, and ships it to 98 million properties with a confidence score attached.
What to Do If You Are Building a High-Performance Home
Document everything. Document it in formats the appraisal ecosystem can ingest. This is tedious work that should not be your job, but it is, and pretending otherwise costs you $42,000 at closing.
Get a HERS rating. It is the closest thing to a universally recognized performance metric for residential energy efficiency. A score of 45 means 55 percent less energy than a code-built reference home; every score is registered in the RESNET database, verifiable by any appraiser who bothers to look. Without one, your spray foam and your ERV and your triple-pane windows are just line items on an invoice that the appraiser may or may not glance at, probably will not understand, and almost certainly cannot price, because the training that would have taught them how to price it is the same training that fewer than 5 percent of their profession has completed.
Use the Addendum. Fill it out yourself if you have to. Include performance certifications, HERS score, estimated annual energy savings in dollars, incremental cost of each green feature relative to code-minimum alternatives. Hand it to your buyer's lender with a cover letter requesting it be provided to the assigned appraiser. No lender is required to pass it along. But if the appraisal comes back low and the lender refused to forward documented performance data, that is a conversation worth having.
Write a contract clause. Belcher's approach is unusual but entirely legal, and if you are building homes that cost $15,000 to $65,000 more than code-minimum because of performance features, protecting that investment in the sales contract is not paranoia but project management.
Push your MLS. Green fields exist in most systems but almost nobody fills them in. If your listing agent does not populate them, the next buyer's appraiser will pull comps that show your home as identical to the code-minimum build down the street, and the AVM will learn from that transaction what it learned from every transaction before it: that green features do not affect price, because the only evidence it can see says they never have. One populated field in one closed transaction creates one data point, which is not much but more than zero.
Limitations
We do not have access to ATTOM's training data, feature weighting, or methodology beyond its May 2026 press release and white paper summary. Our fewer-than-5-percent appraiser education figure comes from Sandra Adomatis, a respected but individual expert source; no federal agency publishes comparable data. Congress has not enacted the GREEN Appraisals Act, and we cannot assess its practical impact. MLS green data capture rates vary significantly by market, and some regions may have substantially better coverage than the national average. Our nondiscrimination and disparate-impact analysis does not reflect an established regulatory position. Our $42,000 green premium example uses representative costs from industry sources; yours will vary by region, contractor, and specification.