The points question tends to arrive at the worst possible moment — after the borrower has seen a rate they don't love, while they are already stretched by the cash needed to close, and usually phrased as "just tell me what you'd do." The explanation itself barely changes from borrower to borrower: a point is a fee paid up front to buy down the rate, the lower rate produces a smaller payment, and the two only balance out after enough months have passed. Which is exactly why AI drafting is appealing — keep the explanation clear and patient, stop rebuilding it by hand every time someone asks.
The catch is that a points explainer is unusually easy to get subtly, expensively wrong. It can reassure the borrower they'll "just refinance later," which is a rate forecast in a friendly voice. It can state a break-even the loan officer has no way to promise. Or it can drop in a helpful-sounding line about points being tax-deductible, which is not a mortgage answer at all. The discipline that makes this workflow safe is simple: AI drafts the explanation and the shape of the comparison; the numbers stay on the Loan Estimate, the rate forecast stays out of the draft, the decision stays with the borrower, and the tax question goes to their tax professional.
The three traps in AI points-and-buydown communication
There are three ways this goes wrong, and all three come from letting the draft's fluency reach past the explanation into a forecast it can't make, a number it can't know, or a field it doesn't belong in.
The first is predicting where rates go. It rarely shows up as an outright forecast. It shows up as reassurance — "and if rates come down you can always refinance later," "rates should improve by next year," "marry the house, date the rate." Every one of those is a prediction, and on a points conversation it does specific damage: the entire value of a buydown depends on the borrower keeping the loan, so a casual promise that the rate is replaceable quietly dissolves the trade-off being explained. If rates don't cooperate, the borrower paid up front for a rate they were told they wouldn't have to keep. The safe pattern is to strike rate direction from the draft completely and anchor the decision on what is knowable today: what the buydown costs, what it changes, and how long the borrower expects to hold the loan.
The second is promising the break-even. AI is fluent about arithmetic, so it will state a break-even period, a payment difference, or a total saving with total confidence — and the borrower will read that as a quote from their loan officer. But the real break-even depends on what the lender is actually charging for the buydown on this file, how far the rate actually moves, and how long the borrower will really keep the loan, which is a fact only the borrower has. The safe pattern is to keep every figure off the draft, explain the comparison as a structure — cost now against a lower payment later, measured against how long they'll stay — and point the borrower to the numbers on their own Loan Estimate to run it. There's an advertising dimension here too: a message that quotes a rate, a points cost, or a payment can trigger disclosure obligations under the applicable advertising rules, which is a compliance review, not a drafting decision.
The third is answering the tax question. "Are points deductible?" sounds like a factual question, so AI answers it — confidently, and sometimes with the year-paid-versus-amortized distinction attached. Whether points are deductible and how depends on the borrower's own situation and on rules that change, and a borrower who hears "they're deductible" from their loan officer may decide on that basis. The safe pattern is one neutral line that names the possibility and routes it: the treatment of points depends on the borrower's circumstances and is a question for their tax professional — no assertion, no "usually," no estimate of the benefit.
There's one clarification worth putting up front, because it is the single thing borrowers most often have backwards: a point is a cost, not a contribution. It is not part of the down payment, it doesn't build equity, and it doesn't shorten the loan — it buys a lower rate, and nothing else. Leading with that prevents half the confusion before the comparison even starts. The close second: a permanent buydown paid with discount points is not the same thing as a temporary, often seller- or lender-funded rate reduction that steps back up after an initial period. Blurring the two is how a borrower ends up surprised by a payment they were never told was temporary.
"Write an encouraging email telling my buyer that paying two points is a no-brainer, that they'll make the money back in under three years, that the payment savings are worth it, that points are tax-deductible anyway, and that if rates drop they can just refinance out of it." In thirty seconds the draft has quoted figures that belong on a Loan Estimate, promised a break-even, given tax advice, and forecast the rate market — four confident errors in one upbeat note.
You confirm what the lender is actually offering; AI wraps it in a calm explainer that says a point is a cost rather than a contribution, distinguishes a permanent buydown from a temporary one, lays out the trade-off as a structure the borrower fills in from their own Loan Estimate, names how-long-you'll-keep-the-loan as the deciding factor, and routes the tax question. Fast, clear, honest — and it forecasts nothing and promises nothing.
A points or buydown message has three sharp edges. The rate forecast isn't yours to make: strike "you can refinance later," "rates should come down," and every softer variant — the decision gets framed on what is knowable today. The break-even isn't yours to promise: the figures live on the borrower's Loan Estimate and the answer depends on how long they will really keep the loan — explain the comparison, state no numbers. The tax answer isn't yours to give: name the possibility in one neutral line and route it to their tax professional. The rule for this workflow: AI drafts the explanation; you keep the forecast out, the numbers on the Loan Estimate, the decision with the borrower, and the tax question with their preparer. Your company's advertising and communication policy and the applicable rules govern.
Where AI actually helps — and where it must not
Explaining what a point actually buys — AI helps
"A point is money paid at closing to lower your rate — it isn't part of your down payment and it doesn't build equity" is the clarification that prevents most of the confusion, and AI writes it well. It just has to stop at the concept and leave the cost and the rate difference to the Loan Estimate.
Laying out the trade-off as a structure — AI helps
Describing the comparison — money now against a smaller payment later, and the point at which the two meet — is nearly identical every time and reads better in plain language. AI is good at this, provided it presents the structure the borrower fills in and never states the figures or the break-even itself.
Naming the deciding question — AI helps
The question that actually settles it is "how long do you expect to keep this loan?" AI is good at putting that question at the centre of the explainer, respectfully and without pressure, as the borrower's own input rather than something the loan officer assumes for them.
Distinguishing a permanent buydown from a temporary one — AI helps, carefully
Explaining that discount points buy a rate for the life of the loan, while a temporary buydown reduces the payment for an initial period before stepping up, is a genuinely useful clarification AI can draft — as long as it describes the mechanism generally and never invents the structure, the funding source, or the step-up schedule on a specific file.
Stating the rate, the cost, the payment, or the break-even — the Loan Estimate owns this
Every figure in this conversation belongs on the borrower's own Loan Estimate or written quote, not in a drafted message — and quoting rates, points costs, or payments in a marketing message can trigger disclosure obligations. AI must describe the comparison and route the borrower to their document for the numbers.
Forecasting rates or recommending the decision — nobody owns this
No draft should predict rate direction, imply the borrower can refinance out of the decision later, or conclude that points are or aren't worth it. The first two are forecasts nobody can make; the third is a decision the borrower owns once they know the trade-off and their own timeline.
Answering whether points are deductible — the borrower's tax professional owns this
The tax treatment of points depends on the borrower's circumstances and on rules that change. AI must route the question rather than answer it — one neutral line, no "usually," no estimate of the benefit.
What to settle before you draft the message
A safe points message is built from facts you've already confirmed — not from what the AI decides sounds complete. Have these settled before you generate anything:
- What the lender is actually offering — the real buydown structure available on this file, so the explainer describes a product that exists rather than a generic one the AI assumes. The draft explains a choice you've confirmed.
- Permanent or temporary — whether the conversation is about discount points that buy the rate for the life of the loan or a temporary reduction that steps up later, and who would fund it. Blurring these is the most consequential error in this whole conversation.
- Where the numbers live — that every figure routes to the borrower's Loan Estimate or written quote, and that no rate, cost, payment, or break-even appears in the draft. Decide this before you write, not while editing.
- How you'll handle the timeline question — that "how long do you expect to keep this loan?" is asked, not assumed, and that the borrower's answer drives the decision rather than your estimate of it.
- Your advertising policy — what your company allows in a written message that touches rates, costs, or payments, and what has to go through compliance review before it goes out.
The points-explainer workflow — step by step
Confirm what's actually on the table
Pull the real buydown option from the lender's pricing for this file, and note whether it's permanent or temporary. Everything downstream describes this, not a generic version.
Write the frame before you open the tool
Decide in one sentence what the message is: an explanation of what a point buys and what the trade-off depends on, ending with a question about the borrower's timeline. Not a recommendation.
Draft with explicit prohibitions in the prompt
Tell the tool up front: no rates, no costs, no payments, no break-even period, no rate forecast, no refinance-later reassurance, no tax answer, no recommendation. Prohibitions in the prompt are cheaper than corrections in the edit.
Run the no-forecast, no-figure scan
Have the tool flag — not rewrite — anything in the draft that states a number, predicts rate direction, promises a break-even, answers the tax question, or recommends the decision. Then fix those yourself.
Route the numbers and the tax question explicitly
Add the two routing lines by hand: the figures are on the borrower's Loan Estimate, and the tax treatment is a question for their tax professional. These are the lines you never let the tool improvise.
Send it with your identifying details, and keep the copy
Your name, company, and NMLS ID belong on the message. Save the version you sent — if the borrower later says they were told points were a sure thing, the drafted text is the record.
Prompts that keep the draft inside the line
Write a short, plain-English explanation for a borrower asking whether they should pay discount points. Rules: - Lead with the clarification that a point is a COST paid at closing to buy a lower rate. It is NOT part of the down payment, it does NOT build equity, and it does NOT shorten the loan. - Explain the trade-off as a STRUCTURE: money paid now against a lower payment later, and a point at which the two balance out. - Do NOT state any rate, points cost, monthly payment, dollar amount, or break-even period. Say the figures are on their Loan Estimate. - Do NOT predict where rates are going, and do NOT suggest they can refinance later if rates drop. - Do NOT say whether points are or are not worth it. That is their call. - End by asking how long they expect to keep the loan, because that is what decides it. - Warm, patient, plain. No hype.
Write a brief explanation of the difference between a permanent rate buydown paid with discount points and a temporary buydown. Rules: - Explain generally: discount points buy a rate for the life of the loan; a temporary buydown lowers the payment for an initial period and then steps up. - Make clear the temporary payment is temporary, and the borrower should plan for the full payment. - Do NOT invent the structure, the step-up schedule, the funding source, or any figure for this borrower's file. - Do NOT recommend either option. - Neutral, clear, specific about the mechanism only.
Review this draft. Do NOT rewrite — just flag. [paste the message] Flag anything that: - states a rate, points cost, monthly payment, dollar amount, percentage, or break-even period instead of routing to the Loan Estimate; - predicts rate direction, or suggests refinancing later if rates drop; - concludes that paying points is or is not worth it; - answers whether points are tax-deductible instead of routing the question to a tax professional; - blurs a permanent buydown together with a temporary one. Return a plain list of what you found and where. Do not add figures, forecasts, or recommendations.
"Hi Marcus — good question, and worth taking a minute on. A discount point is money you pay at closing to buy a lower interest rate. The part that trips people up: it isn't part of your down payment, it doesn't build equity, and it doesn't shorten your loan. It buys a lower rate, and that's all it does."
"So the trade-off is straightforward to describe, even though the answer is personal: you pay something up front, and in exchange your monthly payment is smaller. Those two only balance out after you've held the loan long enough for the smaller payments to add back up to what you paid at closing. Before that point, you're behind; after it, you're ahead. The exact cost and the exact rate difference for your file are on the Loan Estimate I sent — run the comparison against those numbers, not against a rule of thumb."
"Which means the question that actually decides it isn't really about the loan — it's about you: how long do you expect to keep this one? If you're confident you'll be in it a long while, the up-front cost has time to work. If there's a real chance you'll sell or move sooner, it may not. I'm not going to tell you rates will or won't move, because I don't know, and I'd rather you make this call on what we can actually see today. One more thing: people often ask whether points are deductible — that depends on your situation, so please check with your tax professional rather than taking my word for it. Happy to walk through the Loan Estimate line by line whenever you like. — Dana Ruiz, Loan Officer, Example Home Lending · NMLS ID 000000."
Tools that work well for drafting points and buydown communication
Every points message lands in a moment where the borrower wants a straight answer to "what should I do?" Before you send, confirm that no rate, points cost, payment, dollar amount, or break-even period appears in the draft and that the figures route to the Loan Estimate, that nothing predicts rate direction or offers refinancing later as a safety net, that the draft explains rather than recommends, that the tax question is routed rather than answered, and that a permanent buydown is not blurred together with a temporary one. Where a written message touches rates, costs, or payments, your company's advertising policy and the applicable disclosure rules govern — that is a compliance review, not a drafting choice. AI tools don't know your pricing, can't forecast the market, and will happily sound decisive to seem helpful. That judgment is yours.
A note on running points conversations like a professional
The reason loan officers automate this explainer is friction — the same patient explanation of what a point buys, the same walk through the trade-off, the same gentle correction of the belief that points are somehow part of the down payment, rebuilt every time a borrower gets nervous about a rate. AI genuinely removes that friction, which matters when someone is deciding whether to hand over money at closing they can't get back. But the moment these drafts are free to mass-produce, the tempting shortcuts are the ones that turn an honest explainer into a liability: let the tool promise a break-even to sound concrete, add "you can always refinance later" to sound reassuring, or confirm the deduction to sound complete. Each one trades a message that keeps the borrower informed for one that makes a promise nobody can keep.
Use AI to do what it's good at: turning a real trade-off into a clear, calm, consistent explanation — starting with the fact that a point is a cost rather than a contribution, and that a permanent buydown and a temporary one are different animals. Keep the parts that carry the risk where they belong: the figures on the borrower's Loan Estimate, the rate forecast nowhere at all, the decision with the borrower once they've answered how long they'll keep the loan, and the tax question with their tax professional. The points conversation that protects your borrower is the one that explains the trade honestly, states no numbers, predicts nothing, and asks the question only they can answer — and it's your name and NMLS ID, not the tool's, on the signature.
- Because "is it worth paying points?" comes up on almost every rate-sensitive file — and this turns the explanation into a clear draft in minutes, without quoting a figure or promising a break-even.
- To keep the conversation honest, so the borrower hears "here's the trade-off and here's the question only you can answer" instead of a recommendation dressed up as help.
- Because the real risk isn't the writing — it's the rate forecast, the promised break-even, and the tax answer, and this workflow keeps all three out of the draft.
Frequently asked questions
What are discount points, and can a loan officer use AI to explain them?
Discount points are a fee a borrower pays at closing to buy a lower interest rate on the loan. AI is genuinely useful for turning that into a plain-English explainer, because the concept barely changes borrower to borrower. The clarification worth leading with is the one most borrowers miss: a point is a cost, not a contribution. It is not part of the down payment, it does not build equity, and it does not shorten the loan term — it buys a lower rate, and it only pays off if the borrower keeps that loan long enough to recover what they paid. What the draft must not do is state a rate, a points cost, a payment, or a break-even period, predict where rates are heading, or answer whether points are deductible. Use AI for the explanation and the structure of the comparison; keep every figure on the borrower's own Loan Estimate and every tax question with their tax professional.
Can AI tell a borrower whether paying points is worth it?
No. Whether points are worth it turns on facts the AI cannot know and the loan officer cannot assume: what the lender is actually quoting for the buydown, how much the rate actually moves, and — the part that decides it — how long the borrower will really keep that loan. A borrower who sells, refinances, or pays the loan off before recovering the up-front cost has simply paid extra. A draft that concludes 'paying points makes sense for you' has made a decision that belongs to the borrower. The safe pattern is to have AI explain the trade-off and lay out the comparison structure — cost now against the lower payment later, measured against how long they plan to stay — while the borrower runs the arithmetic against the figures on their own Loan Estimate and makes the call.
Is 'you can always refinance later if rates drop' safe to put in a draft?
No — that sentence is a rate forecast wearing the costume of reassurance, and nobody, including the loan officer, knows where rates go. It matters more on a points conversation than almost anywhere else, because the entire value of a buydown depends on the borrower keeping the loan; a casual 'you can just refinance later' quietly undercuts the very trade-off being explained, and if rates do not cooperate, the borrower paid up front for a rate they were told they could replace. The same applies to the softer versions AI reaches for — 'rates should come down,' 'when the market improves,' 'marry the house, date the rate.' The safe pattern is to strike rate direction from the draft entirely and frame the decision on what is knowable today: the quoted cost, the quoted rate difference, and how long the borrower expects to hold the loan.
Are discount points tax-deductible, and can AI answer that question?
Deductibility is a tax question, and it does not belong in a loan officer's draft. Whether points are deductible, and whether they would be deducted in the year paid or spread over the life of the loan, depends on the borrower's own circumstances — the type of transaction, how they file, and rules that change. AI will answer the question confidently anyway, because it reads as a factual question, which is exactly what makes it dangerous: a borrower who hears 'and they're tax-deductible' from their loan officer may decide on that basis. The safe pattern is a single neutral line that names the possibility and routes it — the treatment of points depends on the borrower's situation and is a question for their tax professional — with no assertion, no 'usually,' and no estimate of the benefit.
What should a loan officer never let AI decide in a discount-points or buydown message?
Never let it state a rate, a points cost, a monthly payment, or a break-even period; never let it predict rate direction or suggest refinancing later as the safety net; never let it conclude that points are or are not worth it; never let it answer whether points are deductible; and never let it blur a permanent buydown paid with discount points into a temporary, often seller- or lender-funded rate reduction — they are different products with different consequences when the temporary period ends. Those cross into figures that belong on the borrower's Loan Estimate, a forecast nobody can make, a decision the borrower owns, and a tax question for their preparer. There is also an advertising dimension: a marketing message that quotes a rate, a points cost, or a payment can trigger disclosure obligations under the applicable advertising rules, which is a compliance review, not a drafting choice. The draft's job is to explain what a point buys, name the trade-off honestly, lay out the comparison structure, and route every number to the Loan Estimate.
Get new AI workflow guides for loan officers
We publish new workflow guides for mortgage professionals. No spam, one email when something useful goes live.