Why general AI can get mortgage-note law wrong

A legal answer can sound reasonable and still be wrong on the one detail that changes the outcome.

In our internal mortgage-note benchmark, we found three state-law examples where plain Claude gave a confident but incorrect answer on a decision-sensitive fact. NoteGage's benchmark workflow checked the legal claim against official statute text before the answer was delivered and returned the keyed answer.

This does not make NoteGage infallible. It shows why legal claims in mortgage-note diligence need a verification step, especially when a deadline, formula, or priority rule can change the economics.

Ohio: a $41,000 deficiency and a two-year cutoff

An investor asked whether they could still garnish a borrower after a judicial foreclosure sale left about $41,000 unpaid. In the benchmark, the mortgage was on an owner-occupied one-family residence, the sheriff's sale had been confirmed in March 2024, and the question was asked in October 2026.

Plain Claude said there was no two-year cutoff and that a five-year dormancy rule was the real concern. It told the investor they were “fine for now.”

For a qualifying mortgage on an owner-occupied one- or two-family residence, Ohio Revised Code 2329.08 says the judgment becomes unenforceable as to the remaining deficiency after two years from confirmation of the judicial sale.

On the benchmark facts, that period ran from March 2024 to March 2026. By October 2026, it had already expired.

The practical consequence was not minor. A buyer relying on the wrong rule could spend time and money trying to collect a roughly $41,000 deficiency that, under the statute applied to these facts, had already become unenforceable.

NoteGage's first draft initially leaned the same wrong way. Before the answer was delivered, its law-checking step compared the claim against the official Ohio statute. The statute contradicted the draft, so the answer was corrected.

Official source: Ohio Revised Code 2329.08

North Carolina: an upset-bid number that was $5,640 too low

After a power-of-sale foreclosure, an investor had made a $142,000 credit bid and wanted to know how much another bidder would need to offer to upset the sale.

Plain Claude used the wrong increase formula and calculated a minimum upset bid of about $143,460.

North Carolina General Statutes 45-21.27(a) requires an upset bid to exceed the reported sale price or last upset bid by at least 5%, subject to a $750 minimum increase.

On the $142,000 benchmark example:

• 5% of $142,000 = $7,100
• $142,000 + $7,100 = $149,100

The minimum upset bid was therefore $149,100, not $143,460. That is a $5,640 difference.

That number changes how much protection the investor's winning bid actually has. Someone trying to place an upset bid using the lower figure would also be short of the statutory minimum.

NoteGage checked the rule against the official North Carolina statute, applied the 5% requirement to the $142,000 example, and returned the $149,100 figure.

Official source: N.C.G.S. 45-21.27

New Hampshire: attorney's fees that can be part of the priority amount

A first-mortgage holder asked about a condominium association lien for 14 months of unpaid dues at $275 per month, plus about $2,100 in legal fees.

Plain Claude correctly recognized that roughly six months of regular assessments can receive priority over a first mortgage, but then said the association's legal fees were generally junior to the mortgage.

New Hampshire RSA 356-B:46 I(c) says the priority portion can include regular monthly common assessments unpaid during the six-month period immediately preceding the filing of the memorandum, together with collection costs, including reasonable attorney's fees, when the statutory requirements are satisfied.

In the benchmark example, six months of dues equaled $1,650. The important qualifier is that not every dollar of every association legal bill automatically jumps ahead of a first mortgage. The priority depends on the qualifying assessment period and the statute's delinquency and notice conditions.

But treating the attorney's fees as simply junior can still materially understate the amount that may have priority.

NoteGage checked the official New Hampshire statute, included reasonable collection attorney's fees in the potential priority analysis, and separately told the user to verify whether the required delinquency and pre-filing notice conditions had been satisfied.

Official source: New Hampshire RSA 356-B:46

Why the verification step matters

These three failures were different. One missed a statutory cutoff. One used the wrong bid formula. One remembered the basic priority rule but omitted a component that could sit ahead of the first mortgage.

The common problem was not that the answers sounded reckless. They sounded plausible.

For decision-sensitive mortgage-note law, plausibility is not enough. A model-generated legal claim should be checked against the available official source before it becomes the final answer. If the official text contradicts the draft, the claim should be corrected. If official support cannot be verified, the point should stay labeled as unverified rather than being presented as settled law.

The Ohio example shows why that extra step matters. NoteGage's own first draft initially leaned toward the same wrong answer as plain Claude. The correction happened because the workflow checked the claim before delivery.

Supporting benchmark result

These examples came from an internal mortgage-note Q&A benchmark, not a whole-deal comparison.

Across four internal mortgage-note benchmark rounds, NoteGage Expert's full configuration produced essentially the same number of correct keyed facts as plain Claude (331 vs. 332), while making 7 wrong assertions versus Claude's 34, including 6 critical errors versus 25.

Short form: about the same number of correct keyed facts, with 79% fewer wrong assertions and 76% fewer critical errors.

We treat that as reliability evidence, not proof that NoteGage is already the better expert on every mortgage-note deal.

The practical takeaway

A determined investor can research state law manually. The value of a specialist workflow is not that the law is impossible to find. It is that a model-generated legal claim should not become a final answer merely because it sounds right.

When a deadline, priority rule, or statutory formula can change the economic outcome of a note, the official source should be checked and unresolved conditions should remain visible.

NoteGage provides educational information, not legal advice. Mortgage-note laws are state-specific and fact-specific. Verify legal conclusions with a licensed attorney in the relevant state.