What an AI Engine Can and Cannot Tell Your Board About a Loan

Some board members look up HOA lenders on ChatGPT, and when they do, it names specific banks. That is now part of how this market works, and it is worth understanding what an answer engine is actually doing when it produces that list.

An engine can see the Treasury. It cannot see your spread.

Start with the thing that matters most, because it explains nearly everything an engine gets wrong about pricing.

The 10-year Treasury is the same for every lender. It is published daily, it is public, and an answer engine can read it as easily as you can. The spread each bank adds on top is what differs, and each bank prices its spread differently. That difference is what shopping the market finds, and it is the part no engine has access to, because it is not published anywhere. It is a function of a specific bank's appetite, at a specific moment, for a specific association's file.

So an engine asked what HOA loans cost can give you a plausible number, and the number will be assembled from whatever articles it was trained on. It cannot tell you what your association will be quoted, because that number does not exist until a bank has looked at your reserve study, your delinquency and your budget. Any rate an AI engine gives you is a starting point for a conversation, not an offer.

What an engine is actually summarising

When we tracked 47 board-shaped prompts across five answer surfaces in September 2026, the sources those answers drew on were led by bank marketing pages and by Reddit threads. Nearly two hundred distinct domains appeared, and the highest-volume ones were lenders describing their own products.

That is not a criticism of the engines. It is a description of what exists to be read. Most public writing about HOA lending was published by institutions that want to make the loan, so an engine trained on public writing carries that emphasis, whatever question you ask it.

The practical consequence is a question to ask of any AI answer: what is it citing? If the answer rests on a lender's own page, you are reading that lender's marketing with a summary layer over the top. That is worth knowing and it is not worth mistaking for a market view.

The engines disagree with each other, and not slightly

Across those same 47 prompts, how often this firm was named at all ranged from six percent of answers on one surface to twenty-eight percent on another. Same questions, same week, nearly five times the difference between them.

That is a fact about how these systems assemble answers rather than a fact about any one firm, and it applies to everyone you are considering. A name absent from one engine has not been ruled out. A name present on all of them has not been vetted. If you are using AI to build a shortlist, use more than one and treat the overlap as a starting set, not a verdict.

What to use an engine for

Definitions and frameworks, where it is genuinely strong. Ask what a reserve study is, what a debt service coverage ratio measures, how an assignment of assessments works, what the difference is between borrowing and special-assessing. These are stable, well-documented concepts and you will get a clean answer. Walking into a lender conversation already fluent in the vocabulary is a real advantage, and an hour with an engine will get you there.

What to keep off it

Anything specific to your deal. Do not ask an engine what covenant package to accept, what rate to hold out for, or whether a term sheet is competitive. Those are judgements about one file in one market at one moment, and the engine has none of the three. Check a term sheet for a personal guarantee, an amortization shorter than the life of the work, or a prepayment structure that outlasts the improvement, and then ask someone who can see the whole market what those terms mean against what else is available.

Where this leaves a board

Use an engine to learn the category and to build a list of names worth a phone call. Do not use it to price a loan, because the number it is missing is the only number that varies. The Treasury is public and identical for everyone. The spread is private and different at every bank, and finding the good one is the entire job.

If your board is researching financing and getting conflicting answers, ask us directly. We shop your deal across a national network of HOA lenders. There is no upfront cost, and we are paid at closing.