ChatGPT Prompts for Mortgage Brokers 2026: 15 Ready-to-Use Scripts

A senior loan officer in Phoenix closed 38 loans in August. She never touched a blank page. Every borrower email, every realtor cold-open, every TRID explainer started as a ChatGPT draft she edited in under two minutes. Her team of three does the volume of a team of seven from 2023. This is not a productivity story. It is a margin story, and the brokers who ignore it in 2026 are losing files to the ones who did not.

This guide gives you 15 ChatGPT prompts for mortgage brokers 2026 that you can paste into ChatGPT, Claude, or Gemini today. They cover pre-qualification screening, first-time buyer scenarios, rate-lock timing, refinance breakeven, FHA vs VA vs conventional comparisons, realtor cold-calls, stalled-app follow-ups, credit coaching, TRID disclosure translation, and post-close referral asks. Each one has fill-in slots so you keep control of the borrower-specific detail.

What mortgage brokers can actually automate with ChatGPT

Not every part of loan origination should touch an AI model. Underwriting decisions, DU/LP runs, credit pulls, and any signature capture stay in your LOS. What ChatGPT is good at is the writing layer around those systems: the borrower-facing explanation, the realtor-facing pitch, the internal file note, the follow-up cadence.

Here is where mortgage brokers are getting real hours back in 2026:

• Drafting pre-qual intake emails that ask the right income and asset questions without sounding like a form
• Translating a Loan Estimate into a two-paragraph plain-English summary a first-time buyer can actually read
• Writing rate-lock timing emails that explain float-down risk without practicing law
• Running refinance breakeven math and turning the numbers into a decision-grade one-pager
• Producing side-by-side FHA, VA, and conventional comparison scripts for buyer consultations
• Cold-call and cold-email scripts for realtor partnership development
• Structured follow-up sequences for applications that stalled at conditional approval
• Compliance-aware credit repair coaching outlines you can hand to a borrower without giving legal advice
• Post-close referral asks that do not feel transactional

Feed ChatGPT your state, your lender panel, and your typical borrower profile in a saved system prompt. Every draft after that is 3x more accurate. A generic prompt writes generic emails. A prompt that knows you work Texas non-QM will write Texas non-QM copy.

15 copy-paste ChatGPT prompts for mortgage brokers

Paste each of these into ChatGPT-5, Claude Sonnet 4.5, or Gemini 2.5 Pro. Replace anything in square brackets with your borrower or scenario detail. Read the output before you send. Every one of these prompts assumes you are the licensed originator making the final call.

1. Pre-qualification client screening intake

You are helping me pre-qualify a new borrower for a home purchase mortgage. Draft a friendly but thorough intake email that asks for:
- Gross monthly income and pay structure (W-2, 1099, self-employed, bonus/commission split)
- Total monthly debt payments (auto, student, credit card minimums, child support)
- Liquid assets available for down payment and reserves, plus source (savings, gift, 401k)
- Estimated credit score range and any recent late payments or collections
- Target purchase price, down payment percentage, and timeline
- Whether they are working with a real estate agent

Tone: warm, human, not a form. 180 to 220 words. Sign off as [YOUR NAME], [YOUR TITLE], NMLS #[YOUR NMLS]. Do not promise approval, rate, or program eligibility. Include one line that says any pre-qualification is subject to full underwriting and documentation review.

Borrower first name: [BORROWER FIRST NAME] State: [YOUR STATE] Loan purpose: [PURCHASE / REFI]

2. Loan scenario explanation for a first-time buyer

Write a plain-English scenario walkthrough for a first-time homebuyer who has never seen a Loan Estimate before. Use these numbers:


- Purchase price: $[PRICE] - Down payment: $[DOWN] ([X]%) - Loan amount: $[LOAN] - Program: [FHA / CONVENTIONAL / VA] - Interest rate: [RATE]% - APR: [APR]% - Term: [30-year fixed] - Estimated P&I: $[PI] - Estimated taxes and insurance escrow: $[TI] - Estimated MI or MIP: $[MI] - Total monthly PITI: $[PITI] - Estimated cash to close: $[CTC]

Explain what each line means in one or two sentences. No jargon. No abbreviations without defining them once. End with three questions the borrower should ask themselves before they sign. Do not give tax or legal advice. Do not tell them whether to buy the home.

3. Rate-lock timing email

Draft an email to a borrower who is under contract with a closing date of [CLOSE DATE]. Their current quoted rate is [RATE]% on a [PROGRAM] loan. The lock options available are 15, 30, 45, and 60 days. Market context: [ONE SENTENCE ABOUT CURRENT RATE TREND, e.g. "the 10-year Treasury has moved 12 bps in the last week"].


Explain in plain English: - What locking now protects them from - What floating exposes them to - Approximate cost difference between lock lengths (in bps, do not quote exact numbers unless I give them) - Whether a float-down option is available and what it typically costs

End by asking them to confirm their choice by [DEADLINE]. Do not predict rate direction. Do not tell them what to do. 220 to 260 words.

4. Refinance breakeven analysis one-pager

Build a refinance breakeven summary I can send a borrower. Inputs:


- Current loan balance: $[BALANCE] - Current rate: [RATE]% - Current P&I: $[PI] - New rate quoted: [NEW RATE]% - New P&I: $[NEW PI] - Estimated total closing costs (rolled into loan or paid at close): $[COSTS] - Cash-out amount, if any: $[CASHOUT] - Years remaining on current loan: [YEARS]

Calculate: 1. Monthly payment savings (new P&I minus current P&I) 2. Simple breakeven in months (closing costs divided by monthly savings) 3. Total interest paid on current loan if held to maturity vs new loan 4. Whether they reset their amortization clock and what that means

Present as a short one-pager with a plain-English recommendation framework (not a recommendation). Note that this ignores tax deductibility, opportunity cost of the cash used at closing, and any prepayment penalty. Remind the borrower to consult a tax professional on any deduction impact.

5. FHA vs VA vs conventional comparison script

Write a 3-minute consultation script I can use on a call with a first-time buyer who is eligible for FHA, VA (if veteran), and conventional loans. Compare the three across:


- Minimum down payment - Minimum credit score typically accepted - Mortgage insurance treatment (upfront, monthly, cancellation rules) - Debt-to-income tolerance - Loan limits for [COUNTY, STATE] - Seller concession limits - Typical use cases where each wins

Borrower profile: credit score [SCORE], down payment available [DOWN AS %], veteran status [YES/NO], target price [PRICE].

Write it as spoken language, not a brochure. Include natural pause points where I would ask "does that make sense?" Do not tell the borrower which program to pick. End with two questions I should ask them to narrow the choice.

6. Realtor cold-call opener

Draft a 45-second cold-call opener for a realtor I have never spoken to. My differentiator: [1-2 SENTENCES, e.g. "I close purchase loans in 18 days on average and I answer texts on weekends."]. The realtor's recent activity: [DETAIL FROM MLS / SOCIAL, e.g. "three listings in [ZIP], two under $500k"].


Structure: 1. Name and reason for the call in one breath 2. One specific reference to their business (not flattery) 3. One concrete thing I can do for their next buyer 4. A soft ask for a 15-minute coffee or Zoom

Do not use the words "partner," "synergy," "circle back," or "reach out." Sound like a person, not a script. 100 to 130 words spoken.

7. Realtor cold email (post-open-house)

Write a cold email to a listing agent whose open house I attended yesterday at [PROPERTY ADDRESS]. Subject line max 6 words. Body 90 to 130 words.


Include: - One genuine observation from the open house (staging, price point, buyer traffic) - My close-rate data: [YOUR STAT, e.g. "92% of my pre-approvals close"] - An offer to run a buyer through pre-qual within 24 hours if they refer one - A soft CTA for a coffee, not a meeting

Sign as [YOUR NAME], NMLS #[YOUR NMLS]. No emoji. No exclamation points.

8. Stalled-application follow-up sequence

Design a 4-touch follow-up sequence for a borrower whose application has been stalled at conditional approval for 12 days. Missing items: [LIST, e.g. "updated bank statement, letter of explanation on a $4,200 deposit, homeowners insurance binder"].


Touch 1 (day 0): email, warm, itemized, offers a 10-minute call to walk through anything unclear Touch 2 (day 3): text, one sentence, one item highlighted Touch 3 (day 6): phone call script, 60 seconds, no voicemail template Touch 4 (day 10): email to borrower and realtor together, closing date at risk, clear deadline

Tone escalates from helpful to urgent without becoming rude. Every touch reminds them the file cannot move to clear-to-close until the items are received. Do not threaten. Do not blame.

9. Credit-repair coaching outline (compliance-aware)

Write a coaching outline I can review with a borrower whose middle FICO is [SCORE] and who needs [POINTS] points to qualify for [PROGRAM]. Their tri-merge shows: [SUMMARIZE, e.g. "one 30-day auto late 8 months ago, two credit cards at 78% and 84% utilization, one medical collection for $340"].


Cover, in general educational terms only: 1. How utilization affects scoring and what a healthy target range looks like 2. What paying down revolving balances typically does to a score, and typical timeframes 3. What a goodwill letter is and when borrowers sometimes send one 4. What the FCRA dispute process looks like at a high level 5. What medical collection reporting rules changed in recent years

Do NOT: - Tell the borrower to dispute accurate information - Promise a specific score outcome or timeline - Recommend a specific credit repair company - Provide legal advice about the FCRA, FDCPA, or state credit laws

End with a line telling the borrower to consult a nonprofit credit counselor (NFCC member) or attorney for anything beyond general education.

10. TRID Loan Estimate plain-English translator

Translate this Loan Estimate into plain English for a borrower who has never seen one. Explain, section by section:


- Page 1: Loan Terms box, Projected Payments box, Costs at Closing box - Page 2: Loan Costs (A, B, C), Other Costs (E, F, G, H), Calculating Cash to Close - Page 3: Comparisons box (APR, TIP, 5-year cost), Other Considerations

Borrower reading level: 8th grade. Use short sentences. Define APR, TIP, escrow, and "points" the first time each appears. When a line item is one the borrower can shop for (services they can shop for), say so explicitly.

End with three questions the borrower should ask me before they sign an intent to proceed. Do not give legal advice about the disclosure. Do not tell the borrower whether the loan is a good deal.

Loan Estimate data: [PASTE OR SUMMARIZE KEY FIGURES]

11. Post-close referral ask (30 days after funding)

Write a 30-days-post-close email to a borrower I just closed. Loan program: [PROGRAM]. Property: [CITY, STATE]. Closed on: [DATE].


Structure: 1. One sentence congratulating them on something specific (first mortgage payment about to hit, first month in the home, a detail from our closing conversation) 2. A brief, honest check-in on how they are settling in 3. A short reminder of the 2-3 things I do best (purchase, refi, jumbo, first-time buyer, whatever) 4. A specific ask: "if you know one person shopping for a home this fall, would you introduce us?" 5. My contact info and NMLS

150 to 190 words. No exclamation points. No stock phrases like "it was a pleasure." Sound like I actually remember them.

12. LOE (letter of explanation) draft for a deposit

Draft a letter of explanation the borrower can review, edit for accuracy, and sign. It should explain a $[AMOUNT] deposit on their [BANK] statement dated [DATE].


Source of funds per borrower: [DESCRIPTION, e.g. "reimbursement from brother for shared vacation rental"].

Requirements: - First person, borrower voice - 4 to 6 sentences - States the source, the relationship (if applicable), that the funds are not a loan and do not require repayment (if true), and offers to provide supporting documentation - No boilerplate, no legalese

Remind the borrower that they must confirm every fact is true before signing. This is a legal document.

13. Jumbo scenario handoff to a portfolio lender

Write a warm handoff email to my portfolio lender contact at [LENDER NAME]. Borrower scenario:


- Loan amount: $[AMOUNT] - LTV: [LTV]% - Credit score: [SCORE] - DTI: [DTI]% - Reserves: [MONTHS] months PITI - Employment: [W-2 / SELF-EMPLOYED, YEARS] - Property type: [SFR / CONDO / 2-4 UNIT] - Why it needs a portfolio look: [1-2 SENTENCE REASON, e.g. "K-1 income averaged over 3 years, agency AUS came back Refer"]

Tone: peer-to-peer, no fluff. 120 to 160 words. End by asking whether their guidelines can accommodate the scenario and what documentation they would want to see first.

14. Buyer consultation agenda

Build a 30-minute buyer consultation agenda for a first-time buyer meeting me tomorrow. They have a [SCORE] credit score, [DOWN] saved, and a target price of [PRICE] in [CITY, STATE]. They have not been pre-approved anywhere else.


Agenda structure with time boxes: - Minutes 0-3: introductions, what to expect from the call - Minutes 3-10: their goals, timeline, and any concerns - Minutes 10-20: program overview (FHA, conventional, VA if eligible) at a high level - Minutes 20-27: what documents I will need to issue a pre-approval - Minutes 27-30: next steps and how I communicate during a file

Include 2 open-ended questions I should ask in the first 10 minutes. Do not include a rate quote. Do not include a program recommendation.

15. Compliance-safe social post about rate movement

Write a LinkedIn post about mortgage rates for the week of [DATE RANGE]. Rates moved [DIRECTION] by roughly [BPS] bps on the [PROGRAM] 30-year.


Requirements: - 90 to 140 words - No specific rate quoted (say "in the low 6s" or similar range language only if accurate) - No triggering terms per Regulation Z advertising rules (do not quote a rate or APR without the full required disclosures) - No "call me now," no urgency manipulation - One useful observation about what moved the market (Fed language, jobs report, CPI print) - End with a soft CTA: "if you are within 90 days of a purchase or refi, DM me and I will run your scenario." - Sign with NMLS #[YOUR NMLS] and Equal Housing Lender line

Do not predict where rates go next.

ChatGPT vs Claude vs Gemini for mortgage work

All three major models can run these prompts. They are not interchangeable. Here is what brokers are reporting in 2026 after a year of side-by-side use.

The consensus among originators running all three in 2026: draft in ChatGPT for warmth, verify numbers in Claude, and use Gemini for bulk variants when you need 40 realtor emails personalized to 40 zip codes. No single model wins every task.

You can find hundreds more originator-tested prompts in the PromptSpace prompt library, and the sibling guide for financial advisors covers cross-referral scripts for wealth-management partners.

What ChatGPT cannot do for a mortgage broker

The honest limitations matter more than the wins, because the wins are obvious and the limitations get people in trouble. ChatGPT is a writing assistant. It is not a licensed mortgage loan originator, a lawyer, an accountant, or an underwriter. Treat any output like a first draft from an intern who has read every mortgage textbook but has never sat across from a borrower.

Specific things you should not do with ChatGPT:

• Do not paste a full borrower application, tax returns, W-2s, or a tri-merge credit report into a public model. Even paid ChatGPT plans have data-handling terms that may not meet your GLBA safeguards obligations. Redact or use an enterprise-grade deployment with a signed BAA/DPA and zero retention.
• Do not let ChatGPT quote rates, APRs, or program eligibility to a borrower without a licensed human review. Regulation Z advertising rules apply to every rate you put in writing, and hallucinated numbers are still your license on the line.
• Do not use ChatGPT to give tax advice on mortgage interest deductibility, points, or cash-out proceeds. That is a CPA's job. The CFPB and state regulators care.
• Do not use it to give legal advice about disclosures, contracts, TRID timing cures, or RESPA Section 8. That is a real estate attorney's job.
• Do not present AI-drafted material as your own personal writing in a state that requires disclosure of AI-generated communications. Several states passed disclosure requirements in 2025 and 2026.

RESPA, tax/legal advice, and AI disclosure. Every prompt in this guide is a starting draft, not a finished borrower communication. Under RESPA Section 8, anything you send to a settlement service provider (realtor, title agent, appraiser, insurance broker) is scrutinized for referral-for-value language, and an AI-written cold email is judged by the same standard as one you typed yourself. Do not offer or accept anything of value in exchange for referrals. Never give tax advice or legal advice, even when a borrower asks directly, refer them to a licensed CPA or attorney. Several state regulators (including California, Colorado, Texas, and Utah) have enacted or proposed rules requiring disclosure when AI generated content is used in consumer-facing communications, check your state's current requirements before sending AI-drafted material to borrowers. When in doubt, disclose that AI helped draft the text and you reviewed and approved it. Your NMLS license and your compliance officer both prefer over-disclosure.

Building this into your daily loan-officer workflow

A prompt library that sits in a Google Doc dies in a Google Doc. The brokers getting real leverage from these prompts have built them into the moments where they already work: their LOS, their CRM, their inbox.

A workable daily rhythm for a producing LO doing 8-15 files at a time:

1. Morning file review, 20 minutes. Open every active file. For each one at a status change (application received, conditional approval, clear to close), paste the file summary into ChatGPT and generate the next borrower email or realtor update. Edit, send, log.
2. Midday realtor outreach block, 30 minutes. Use prompts 6 and 7 to produce 5 to 10 targeted realtor touches per day. Track responses in your CRM, not in your head.
3. Afternoon stalled-file sweep, 15 minutes. Anything sitting more than 5 days without borrower response gets prompt 8's touch sequence started or advanced.
4. End of day, 10 minutes. Post-close referral asks (prompt 11) for anything that funded 30, 60, or 90 days ago.

Save your five most-used prompts as ChatGPT Custom Instructions or Claude Projects with your NMLS number, state, lender panel, and typical borrower profile pre-loaded. Every draft after that is 60% closer to send-ready. The setup takes 20 minutes and saves you 4 hours a week for as long as you originate loans.

For firms running more than 3 LOs, consider building a shared prompt library on your intranet with your compliance team's pre-approved language. A junior LO editing an approved template is faster and safer than a junior LO writing borrower emails from scratch. The Mortgage Bankers Association and the NMLS Consumer Access guidelines both cover advertising and communication standards you should benchmark against. Related toolkits for adjacent professionals like CPA firms show the same pattern in a different vertical.

Frequently asked questions

Are these ChatGPT prompts compliant with RESPA?

The prompts themselves are drafting tools, not communications. Compliance depends on what you send. RESPA Section 8 prohibits giving or accepting anything of value for referrals of settlement services, so a realtor cold email that offers a gift, a kickback, or a co-marketing arrangement priced below fair market value is a problem regardless of who or what wrote it. Have your compliance officer review any template you plan to send at scale.

Can I paste a full 1003 or tri-merge credit report into ChatGPT?

Not into a consumer ChatGPT account. Your GLBA safeguards obligations require you to control where nonpublic personal information (NPI) flows. Use an enterprise deployment with a signed DPA and zero data retention, or redact borrower name, SSN, and account numbers before pasting. Most brokerages in 2026 have moved to enterprise ChatGPT, Claude Enterprise, or a private-model setup for exactly this reason.

Do I have to disclose that I used AI to draft a borrower email?

It depends on your state. As of 2026, California, Colorado, Texas, and Utah have rules that touch AI disclosure in consumer-facing communication, and several other states have pending legislation. The safest default is to disclose when the AI-drafted content is substantial and unreviewed. If you reviewed and edited the draft, most current rules treat it as your writing. Check with your compliance officer or state regulator.

Will ChatGPT quote accurate mortgage rates?

No, and you should not let it. Any rate a model produces is either invented or based on stale training data. Rates come from your pricing engine, your investor rate sheets, or your LOS. Use ChatGPT to explain a rate you have quoted, never to generate one.

Which model handles Loan Estimate and Closing Disclosure comparisons best?

Claude Sonnet 4.5 as of late 2026 handles side-by-side LE vs CD analysis most reliably because its long-context window keeps every line item in view and it flags material changes without hallucinating figures. ChatGPT-5 is close. Always verify tolerance-cure math yourself against the actual disclosures.

Can I use ChatGPT to write my NMLS continuing education essays?

No. NMLS-approved CE providers require your own work, and using AI to complete coursework can be a licensure issue. Use ChatGPT to summarize a topic you are studying, not to submit content on your behalf.

How do I stop ChatGPT from sounding generic?

Two moves. First, load a Custom Instruction or Claude Project with your voice, your state, your typical borrower, and your lender panel. Second, add a line to every prompt: "do not use the words robust, leverage, unlock, or partner." Bland language usually comes from the model reaching for filler when your prompt is too abstract. Give it real detail, get real copy back.

Is it worth paying for ChatGPT Plus or Claude Pro if I only close 4-6 loans a month?

Yes. The subscription cost is roughly the price of one borrower gift basket per month and pays for itself the first week you use it to draft realtor outreach and stalled-file follow-ups. Small-volume brokers actually benefit more per file because they cannot afford a full-time processor doing the same writing.

Can these prompts help me pass state licensing exams or SAFE?

They can help you study, not test. Use ChatGPT to explain a concept, quiz yourself, or work through a sample TILA calculation. Do not rely on any model for regulatory citations, use the NMLS resource center and your state statute directly. Model outputs get citations wrong often enough that trusting them on the SAFE exam is a bad bet.

What is the single biggest mistake mortgage brokers make with ChatGPT?

Sending unedited output. Every horror story starts with an LO who copy-pasted a draft into an email and hit send. The prompts in this guide save you 80% of the drafting time. The remaining 20% is you, reading every word, applying your license and your judgment, and taking responsibility for what leaves your inbox. That last step is not optional.

Where to go from here

Pick three prompts from the list above. Load them into ChatGPT with your NMLS number, your state, and your typical borrower profile pre-filled. Use them on live files this week. Track how much time you save and how many of your drafts you edit before sending. Most brokers land somewhere between 3 and 6 hours a week returned, and their borrower and realtor communication gets more consistent, not less personal.

Then expand. Add the credit coaching outline, the LE translator, and the referral ask. Build a habit of running every new borrower communication through a prompt before you write it from scratch. Six months in, you are the LO closing 38 files a month with a team of three. The tooling is not the differentiator anymore, the discipline is.

Browse the full PromptSpace prompt library for adjacent workflows, and check the PromptSpace developer tools if you want to wire these prompts directly into a CRM or Zapier workflow. Free foundational docs on model capability and safety live at OpenAI's usage policies and Anthropic's acceptable use policy, both of which every producing LO should read once.