Your Home's Energy Score Was Generated From a Photo and a Tax Record. The Buyer Paid $20,000 Extra For It.

In January 2026, Berkeley, California started requiring every single-family home seller to obtain a Home Energy Score before listing. Sellers must either complete energy upgrades, prove they already have a heat pump, or put $5,000 in escrow. Portland has required scores since 2018. Massachusetts has a bill in committee. At least nine other cities and states have active programs or pending legislation.

A Home Energy Score costs $150 to $500 for an on-site assessment. An AI-generated estimate from public records and satellite imagery costs nothing.

Guess which one is gaining market share.

What the Algorithms Actually Measure

Two companies dominate the automated home energy estimate market. ClearlyEnergy and UtilityScore draw from tax assessor records, utility rate databases, and public building data to generate energy cost predictions for any home nationwide. No visit required. No assessor. No $437 invoice.

Researchers at the University of Notre Dame pushed further. In a 2024 Building and Environment study, they trained a convolutional neural network on Google Street View images to predict household energy expenses across 300,000 Chicago homes. The system analyzed window size, window type, and shading coverage from exterior photos alone. Accuracy: 74%.

Seventy-four percent sounds reasonable until you realize it means the system is significantly wrong about one in four homes.

How Wrong Is Wrong Enough to Matter?

Rocky Mountain Institute conducted the most rigorous comparison available. They analyzed nearly 8,000 homes across 27 states, matching automated estimates from ClearlyEnergy and UtilityScore against DOE Home Energy Score on-site assessments.

Nearly three-quarters of homes fell within 30% of the on-site score. About half were within 20%. Only a quarter were within 10%. That last number is the one that matters for legal purposes. If your city requires an energy score on every listing, and three out of four automated scores deviate from reality by more than 10%, you have embedded a systematic error into the real estate transaction.

A separate UCL study of 1,374 British households using smart meter data found an even more troubling pattern. Energy Performance Certificates, the UK equivalent of Home Energy Scores, overpredicted energy use by 8% for moderately efficient homes (Band C) and by 48% for inefficient homes (Bands F and G). The worse the home, the more the score flatters it.

Read that again. Inefficient homes, the ones where accurate scoring matters most for buyers, are the ones where the scores are most wrong.

How Scores Become Dollars

An analysis cited by Portland's Bureau of Planning and Sustainability, drawing on more than 20 international studies, found that green-certified homes sell for up to 4% more than comparable properties. On a $500,000 Portland home, that is $20,000. On a $1.2 million Berkeley home, that is $48,000.

This premium is the entire policy rationale. Portland explicitly argues that energy scores create "a value that can be recognized by the market" and help sellers "recoup" efficiency investments. Berkeley's BESO ordinance goes further, requiring upgrades or escrow deposits tied to the score.

I call it "phantom value" because it represents the portion of the green premium that rests on inaccurate scoring. If a home scores an 8 out of 10 on its energy rating but actually performs like a 6, the buyer paid for efficiency that does not exist. On the utility bill, that gap shows up as $300 to $600 per year in higher-than-expected costs. Over a 30-year mortgage, the phantom savings compound to $9,000 to $18,000 in net present value, assuming a 3% discount rate.

Who Is Liable?

Nobody. That is the problem.

Portland's ordinance requires the score to appear in the MLS Public Remarks section and a printed copy placed "in an accessible and central location in the home where buyers will see it, such as the kitchen counter or dining room table." But the ordinance does not specify what happens when the score is materially wrong.

Standard real estate disclosure law covers known material defects. A seller who knows the roof leaks and fails to disclose it faces liability under state disclosure statutes. Virginia's statute provides a one-year window from receipt of disclosures. California's Transfer Disclosure Statement covers defects the seller knows about or should reasonably know about.

Energy scores occupy a different category entirely. A seller in Portland is required to hire a qualified assessor and disclose the result. If the assessor generates an inaccurate score, the seller disclosed in good faith. The assessor arguably performed to the standard of the DOE's scoring methodology. The buyer relied on a number that the law required but nobody guaranteed.

Now layer in automation. If ClearlyEnergy generates a score of 8 and an on-site assessment would have produced a 5, who is the buyer's claim against? The algorithm vendor has no direct relationship with the buyer. The seller used a tool the market accepted. The city mandated disclosure but did not mandate accuracy thresholds.

Zillow faced a similar question with its Zestimate in 2017. A homeowner sued, claiming the algorithm had undervalued her property. A federal court dismissed the case, holding that the Zestimate was an opinion, not an appraisal. That precedent cuts both ways here. If an AI energy score is an "opinion," it cannot be the basis for mandated disclosure. If it is a regulated metric, it needs accuracy standards. The current framework gives it the authority of a regulation with the accountability of a blog post.

What Berkeley Got Right and Wrong

Berkeley's BESO ordinance is the most aggressive residential energy disclosure policy in the country. It does not just require a score. It requires action.

Sellers must hire from a registered assessor list, earn at least six upgrade credits (heat pump water heater, solar panels, insulation), or deposit $5,000 in escrow with the city. Buyers who accept the deferral have two years to complete upgrades and reclaim the deposit.

What Berkeley got right: they tied the score to on-site assessment exclusively. No remote estimates. No AI shortcuts. A registered assessor walks through the building. The score feeds into a decision framework with real financial consequences. That is how disclosure should work.

What Berkeley got wrong: they assume the on-site scoring methodology itself is accurate. The UCL data says it is not. On-site assessments using the DOE's scoring model are better than remote estimates, but they still rely on standardized assumptions about occupant behavior, local climate data, and equipment performance that may not reflect reality. A blower door test measures infiltration. A HERS rater models energy consumption. Models produce estimates, not measurements.

What You Should Do

If you are buying in a disclosure-required city: Do not treat the energy score as a guarantee. Request the full Home Energy Score Report, not just the number. Look at the individual component ratings. If the score claims high efficiency but the home has single-pane windows, no insulation documentation, or a 15-year-old HVAC system, the model may be wrong. Budget $300 to $500 for your own independent energy audit with blower door test before closing. On a $500,000 purchase, that is 0.1% insurance against a 4% phantom premium.

If you are selling: Insist on a qualified on-site assessor, even if your city allows automated estimates. An on-site score is harder to challenge after closing. Keep the assessor's detailed report. If a buyer later claims the home underperforms, the report is your evidence of good-faith disclosure.

If you are a builder selling new construction: Get a HERS rating during construction, when the rater can inspect insulation before drywall closes the walls. A pre-drywall HERS inspection is more accurate than any post-construction score and costs $300 to $800 depending on home size. Include it in your marketing as a verifiable performance metric, not just a number on a label.

If you are an assessor: Document your methodology. Photograph the components you evaluated. Note where you relied on the DOE model's default assumptions versus observed conditions. If a dispute arises, your documentation is the difference between professional liability and professional negligence.

Strongest Counterargument

Automated scores, even with 20 to 30% error margins, are categorically better than no information at all. Before Portland's 2018 mandate, fewer than 2% of the city's 160,000 single-family homes had any energy rating. The U.S. has approximately 130 million residential units and fewer than 500,000 have ever received a Home Energy Score. At the current pace of on-site assessments, universal coverage would take centuries.

The accuracy purist's position, carried to its conclusion, kills disclosure entirely. And the data is clear that disclosure, even imperfect disclosure, shifts behavior. Portland reports that homeowners who receive scores are more likely to pursue efficiency upgrades. The EU's Energy Performance Certificate program, despite the accuracy problems documented by UCL, has become the foundation of Europe's building decarbonization strategy.

This is a real tension, not a strawman. Demanding perfect accuracy before allowing any disclosure protects nobody. But mandating scores without accuracy thresholds protects nobody either.

What This Analysis Did Not Prove

The RMI study compared automated estimates against DOE Home Energy Scores, which are themselves modeled predictions, not metered energy consumption. Comparing one model to another does not establish ground truth. The UCL study used smart meter data as the baseline, which is closer to reality, but it studied British housing stock under UK climate conditions. Whether those accuracy gaps transfer to US construction is uncertain.

The 4% green premium figure comes from a meta-analysis of 20+ studies worldwide. Premiums vary substantially by market, home type, and economic conditions. In a rising market, premiums may exceed 4%. In a flat or declining market, the premium may disappear entirely.

No US case law directly addresses liability for inaccurate home energy scores. The Zillow Zestimate dismissal is the closest analogue, but property valuations and energy performance ratings serve different legal functions. Until a court rules on an energy score dispute, the liability framework remains speculative.

And my phantom value calculation assumes the entire green premium is attributable to the energy score. In practice, buyers pay premiums for multiple green attributes: solar panels, efficient appliances, modern insulation. Isolating the score's marginal contribution to the premium would require hedonic regression analysis that, to my knowledge, nobody has published for US residential energy scores.