The Appraiser Who Never Shows Up — What Fannie and Freddie's AI Valuation Waivers Mean for Buyers and Sellers in 2026

What a Waiver Is, and What It Is Not

The term of art at Fannie Mae is “value acceptance.” The rename was deliberate: nothing is being waived in the sense of skipped. The lender submits a value — on a purchase, that is almost always the contract price — and the automated underwriting system decides whether to accept it without independent verification.

The machinery underneath is Collateral Underwriter, Fannie Mae’s proprietary database of millions of appraisals submitted by lenders over more than a decade, paired with analytics that score how plausible a submitted value is. Desktop Underwriter queries it, assesses “the reasonableness of the lender-submitted value for the property,” and returns the minimum collateral due diligence it will require. Freddie Mac runs an equivalent system (Automated Collateral Evaluation) against its own appraisal history. These are not the same thing as the consumer-facing estimates on listing portals, though the underlying statistical approach is related — we looked at how accurate those are in AI home valuations in 2026.

Two features of that design matter more than anything else you will read about waivers.

First, the model needs a prior record of the subject property to work from. Fannie Mae’s own guidance is explicit that the system compares the subject address against appraisals already in Collateral Underwriter. If the house has never been appraised in the modern dataset — new construction, a property that has not transacted in decades, an unusual rural parcel — the model has thin ground to stand on, and typically will not offer a waiver. This is why waivers cluster in ordinary tract housing in active markets and are rare at the edges.

Second, the model is evaluating a price, not a property. It has no idea the furnace is dead or the deck was built without a permit. It answers one narrow question: given what the dataset knows about this address and this neighborhood, is the number on the contract within a plausible range? Nothing in that question touches condition.

The Middle Option Nobody Explains

There is a third path between a full appraisal and a pure algorithmic waiver, and it is the one growing fastest. Fannie Mae calls it value acceptance + property data; Freddie Mac calls its version ACE+ PDR.

Here the model accepts the value but requires “a current view of the property to ensure satisfactory condition and eligibility,” in Fannie Mae’s words. Someone goes to the house and collects a standardized dataset — the Uniform Property Dataset — before closing. It captures interior and exterior characteristics, measurements for a floor plan, photographs, adjacent conditions that help or hurt value, and design defects that impair use. Freddie Mac’s guidance names the bedroom-through-a-bedroom problem as an example of exactly the sort of thing the dataset is built to flag.

Who does the collecting is the controversial part. Freddie Mac permits trained independent professionals, licensed or certified appraisers, or appraiser trainees; the non-appraisers must complete mandatory training on property data collection and the dataset, pass periodic criminal background checks, and comply with Property Data Collection Independence Requirements, with the lender responsible for oversight. What they produce is explicitly “not an appraisal or appraisal report and doesn’t involve…opinion of value.” They document. They do not judge.

The cost difference is the entire commercial argument. Freddie Mac puts property data reports at an average of $200 against about $600 for a traditional appraisal — a saving of nearly $400 — and reports closings 12 days faster on purchases and 10 days faster on refinances. The May 29, 2026 trade letter to FHFA described photograph-based hybrid collection as “five times more efficient” than a traditional appraisal.

How Common Is This Actually? The Numbers

Waiver share is frequently exaggerated in both directions, so here is the picture with sources attached.

The American Enterprise Institute’s Housing Center tracks the split monthly. Its February 2026 data, as reported by the Appraisal Institute on May 5, 2026, breaks down this way.

Purchase loans. Freddie Mac: 77.6% traditional appraisal, 19.9% waiver, 2.5% waiver with property data. Fannie Mae: 85.7% traditional, 11.4% waiver, 2.9% waiver with property data. So on a purchase, roughly one loan in seven to one in five skips the appraiser — meaningful, but far from the norm.

No-cash-out refinances. A different world: Freddie Mac 50.6% traditional and 47.6% waiver; Fannie Mae 49.1% traditional and 47.4% waiver. Close to half of rate-and-term refinances close with no appraisal.

Cash-out refinances. Tighter again, as you would expect when the borrower is extracting equity: waivers ran 27.3% at Freddie Mac and 20.5% at Fannie Mae, with another 6.2% of Fannie Mae loans using hybrid alternatives.

The combined figure across both enterprises was 26% in February 2026 — up from the prior month, but well below the March 2021 peak. That peak is the useful anchor. An FHFA working paper by Joshua Bosshardt, William Doerner, and Fan Xu (November 2022) found waiver use rose from under 10% in early 2019 to over 30% by mid-2021. The pandemic refinance boom, not AI sophistication, drove that spike: when rates collapse, refinance volume explodes, refinances are the easiest loans to waive, and the waiver share mechanically rises. The Appraisal Institute’s May 2026 analysis notes the same mechanism in reverse, observing that no-cash-out waiver share has “generally moved inversely with mortgage rates.”

Which is the first practical implication of a 7.28% October: with rates at the top of the cycle, refinance volume thins and the mix shifts toward purchases, where waivers are far less common. The headline waiver percentage will likely drift down in coming quarters for reasons having nothing to do with policy or model quality.

Purchase waivers, meanwhile, have been climbing on policy. FHFA announced on October 28, 2024 that the maximum loan-to-value ratio for purchase loans eligible for appraisal waivers would rise from 80% to 90%, and for inspection-based waivers from 80% to 97%. Then-Director Sandra Thompson said the change would let “more borrowers, particularly first-time and low- to moderate-income borrowers, benefit from cost savings and reduced closing times.” The effect was immediate and visible: the Appraisal Institute reported in December 2025 that after Fannie Mae opened the 80%–90% CLTV band in the first quarter of 2025, waiver use within that band jumped from roughly 2% in February 2025 to roughly 17% by September 2025.

The point of listing all of that: waiver availability is a policy dial, and it has been turned toward “more” three times in two years.

What a Waiver Actually Saves You

Be careful with the savings claims, because three credible sources give three different numbers and they are measuring different things.

Angi puts the 2026 national average cost of a home appraisal at $359, with most homeowners paying $314 to $425, based on 7,355 verified projects. Freddie Mac uses roughly $600 for a traditional appraisal when comparing against its $200 property data report. The FHFA working paper estimated waivers cut appraisal costs by $300 to $700 and closing times by seven to ten days. Angi measures consumer-ordered appraisals; what appears on a closing disclosure is usually higher, because it runs through an appraisal management company that takes a cut.

The honest range, then: roughly $350 to $700 in cash and seven to twelve days of calendar time. The time may matter more than the money. In a market where the typical home went under contract in 31 days in August 2026 (National Association of Realtors, released September 10, 2026), a ten-day swing in closing timeline is real leverage with a seller choosing between offers.

At the scale of the whole system the money adds up. As of FHFA’s October 2024 announcement, Freddie Mac reported its alternative valuation tool had saved borrowers $1.63 billion in appraisal fees, and Fannie Mae reported its alternatives had saved borrowers more than $2.5 billion since early 2020.

Why the Policy Is Moving Now

Executive Order 14393, “Promoting Access to Mortgage Credit,” signed March 13, 2026, is the clearest statement of federal direction. Section 6(a) directs financial regulators to consider modernizing appraisal regulations to expand the use of “alternative valuation models, desktop and hybrid appraisals, and artificial intelligence valuation tools,” alongside simplifying appraiser qualification requirements and reducing appraisal requirements for low-risk transactions. Section 4 directs the FHFA Director to accelerate collateral and valuation processing through standardized data and digital documentation.

Industry promptly asked for more. The May 29, 2026 letter to FHFA Director Bill Pulte — signed by eight groups including the Broker Action Coalition and the Housing Policy Council, and reported by National Mortgage News on June 1, 2026 — asked FHFA to expand borrower- and agent-captured scanning for hybrid valuations, to raise the property value ceiling for value acceptance from $1 million to $2 million for low-risk properties (one signatory argued there should be no limit and “the data should speak for itself”), and to give appraisers limited access to the enterprises’ appraisal dataset through the Uniform Collateral Data Portal.

Running against that current is one real constraint. The interagency Quality Control Standards for Automated Valuation Models rule, issued June 24, 2024, took effect October 1, 2025 — one year ago this week. It requires institutions using AVMs in covered credit decisions to maintain controls designed to ensure a high level of confidence in estimates, protect against data manipulation, avoid conflicts of interest, require random sample testing and reviews, and comply with applicable nondiscrimination law. That last factor is the first time model fairness in collateral valuation became an explicit federal compliance obligation rather than a research topic.

The Case Against, and What the Data Says About It

Appraisers have been loud, and some of their argument is well taken.

The strongest version is not about valuation accuracy at all. It is that a property data collector — trained and background-checked, but not an appraiser and explicitly barred from offering an opinion of value — is not a substitute for a professional who has spent years learning what to look at. One appraiser quoted in the trade press put it plainly: “We’re out there, on-site, catching things that algorithms and automated valuations can’t even come close to understanding.” Critics writing in Appraisers Blogs objected to inspection “by an Uber driver or unlicensed data collector, rather than a thorough, professional appraiser,” and reported that former FHFA Director Mark Calabria called the LTV expansion “dumb & irresponsible.” The broader fear is 2008 again: that automated valuation at high LTV inflates prices and the system discovers the error only on the way down.

The weaker version is the claim that waived loans are performing badly. The available evidence does not support it. The Appraisal Institute’s December 2025 analysis reports that loans with waivers show lower average default rates, which it attributes to tight credit overlays and careful model governance rather than to the waiver itself. The FHFA working paper found waivers were more common on rate-and-term refinances (41% higher likelihood), loans serviced by nonbanks (10% higher), and higher-credit-score borrowers. On prepayment, it found a 6.7 percentage point raw increase in conditional prepayment rates at 12 months of age during 2020, but only 1.5 percentage points after controlling for observable loan and borrower characteristics — about 78% of the raw association was explained by other factors.

Read those two findings together and you get the correct interpretation. Waivers are currently going to the safest loans in the book. That is a reason to be relaxed about today’s performance data and a reason to be cautious about extrapolating it. The default statistics describe a population selected for low risk. They say very little about how the models would perform on loans at 97% LTV, in a falling market, at scale — which is precisely the direction policy is expanding. “Waived loans default less” is true and is not the same claim as “waiving appraisals is safe.”

What This Means If You Are Buying

You do not choose whether you get a waiver. The automated underwriting system offers one or does not, and your lender passes it along. But you do choose whether to accept it, and you can order an appraisal yourself regardless.

The decision turns on a single question: in this specific transaction, is there independent evidence that the price is right?

Accept the waiver without much worry when the house is conventional, the neighborhood has recent and similar arm’s-length sales, your agent has walked you through comparables you have actually seen, and your down payment is large enough that a modest valuation error does not leave you underwater. Market conditions help here: August 2026 had 1.62 million existing homes for sale, a 4.9-month supply (NAR, September 10, 2026). That is not a frenzy, and in a balanced market contract prices are less likely to be inflated by bidding dynamics — which is exactly when the model’s core assumption, that the agreed price is plausible, is most defensible.

Think harder, and consider paying for an appraisal anyway, when any of these apply: the property is new construction or has not sold in a long time, so the dataset is thin; it is unusual for its area in size, lot, or condition; you are buying in a market where prices have been falling, because the model is trained on history and history is stale on the way down; the transaction is not fully arm’s-length, such as a purchase from family or a builder with incentives rolled into the price; you are stretching to a high LTV, where a 5% valuation error is the difference between equity and none; or you simply cannot assemble comparable sales that make the price make sense. A few hundred dollars to check a $429,100 decision — the August 2026 national median existing-home price (NAR) — is cheap insurance.

And the point that cannot be repeated too often: a waiver is not an inspection, and neither is a property data collection. The data collector documents conditions; they do not evaluate systems, test anything, or tell you what a repair will cost. If your lender does not require an appraisal, that changes nothing about whether you should hire your own inspector. If anything it raises the stakes, because one of the two sets of professional eyes on the house has just been removed. Our home inspection playbook for 2026 buyers covers what inspectors actually find and how to negotiate on it.

What This Means If You Are Selling

For a seller, the appraisal has always been the step where a deal quietly dies — the buyer’s lender values the house below the contract price, the buyer cannot or will not bridge the gap, and you are relisting. A waiver removes that risk entirely.

So when you are weighing offers, ask the buyer’s agent whether the loan has an appraisal waiver. An offer with one carries no appraisal contingency risk and will typically close one to two weeks sooner. Against a marginally higher offer that still has to clear an appraisal, that is a genuine consideration, not a technicality — especially if you are buying your next home on a timeline.

The flip side, and it is a real one: a waiver means the model accepted the price. It is not an endorsement. If you have priced above the comparable sales and a waiver comes through, the model has not validated your number — it has declined to question it, on a transaction it considers low-risk. Sellers who read a waiver as confirmation that the house is worth the asking price are reading something into it that is not there. The pricing discipline described in our 2026 home seller’s playbook still applies.

The Honest Bottom Line

Automated collateral valuation is the most mature application of AI in residential real estate, and on the evidence available it is working as designed. The loans it touches perform well, and it saves real money and real time. The enterprises have so far been conservative about where they deploy it — one in seven to one in five purchase loans, concentrated in the most ordinary properties and the strongest borrowers.

The reason to pay attention anyway is that every force acting on this system pushes one way. The March 2026 executive order asks for more AI valuation; the industry is asking FHFA to double the value ceiling and expand photo-based collection; each LTV expansion so far has been followed within months by a visible jump in waiver use. The performance data everyone cites comes from a deliberately low-risk population, and it will stay reassuring right up until the population changes.

None of that makes a waiver a bad deal today. It does mean the one party with an unambiguous interest in the price being right is you. The lender wants the loan to perform, which is not the same as wanting you to pay the correct amount; the model is checking whether your price is plausible, not whether it is wise. When the dataset is thick and the market is liquid, those two questions have nearly the same answer and the waiver costs you nothing worth keeping. When the property is unusual, the market is soft, or your down payment is thin, they come apart — and the few hundred dollars you saved is the worst trade in the transaction.

Take the waiver when the evidence is already in front of you. Buy the appraisal when it is not. And hire the inspector either way.