Should You Use ChatGPT to Find a Mortgage Broker? Pros and Cons

Los Angeles Mortgage Guide | AI and Mortgages

By Mark Cohen, Cohen Financial Group | #1 Mortgage Broker in the U.S., 2026 Scotsman Guide
Last updated November 1, 2026

Jump to

How It Works | The Pros | Shifting Answers | Hidden Reasoning | Third-Party Sources | 10 Prompts | Vetting a Broker | FAQ

ChatGPT has more than 900 million weekly active users, according to OpenAI’s February 2026 announcement, and homebuyers and homeowners in Greater Los Angeles can ask it a version of the same question: "Who is the best mortgage broker in Los Angeles?" The answer comes back fast, confident, and neatly ranked. Google’s AI Overviews, Gemini, Claude, and Perplexity can be asked the same question.

That is a reasonable place to start and a risky place to finish. This guide uses ChatGPT as the example, and most of it applies to other AI tools too. It covers what they do well, three problems that matter when real money is involved, 10 prompts that get better results, and a way to turn an AI shortlist into a broker you can trust. Cohen Financial Group has placed mortgages for Los Angeles borrowers since 1986.

How does ChatGPT come up with a mortgage broker recommendation?

Depending on your settings and your question, ChatGPT either builds an answer from patterns in its training data or runs a web search and summarizes what it finds. OpenAI’s help center says it can cite web sources when search is turned on. Other AI tools take a similar approach, though the details differ.

Neither path involves a firsthand relationship with any broker. No AI tool has reviewed a broker’s closed loans, spoken with clients, or checked a license. The output reads like a referral from someone who knows the market, but it is a summary of text about brokers, written by other people, some of them the brokers themselves.

The pros: where can AI help you find a mortgage broker?

It explains loan types in plain language

Jumbo loan, bank statement loan, asset depletion loan, HELOC, cash-out refinance. The vocabulary alone slows many borrowers down. AI tools can define each term and describe who the product is built for in seconds, so you arrive at your first broker call knowing the basics.

It helps you write better questions

Ask an AI tool to draft questions for a broker based on your situation, such as self-employment income, a purchase above the conforming loan limit, or a refinance on a rental property. You still need real answers from a licensed professional, but a prepared list keeps you from forgetting what matters.

It gives you a starting list

AI tools can return names of brokers and lenders that serve Los Angeles. Treat every name as a lead, not a recommendation. The list saves time on the first pass, and it is only useful if you check each name afterward.

It can translate paperwork

If you already hold a Loan Estimate, you can ask an AI tool to explain a line item such as origination charges or discount points. Remove your name, Social Security number, loan number, and account details before you paste anything.

AI tools are strongest when they help you think and weakest when they make the judgment. The Consumer Financial Protection Bureau (CFPB) reached a related conclusion in its June 2023 review of chatbots at financial institutions: chatbots handle basic questions, and their usefulness drops as problems get more complex. That report covered customer service bots, not ChatGPT, but a jumbo loan for a self-employed borrower is exactly the kind of complex problem it describes.

Con 1: Ask the same question twice and you can get two different answers

Run the same prompt in two chats and the names, the order, and the reasons can change. SparkToro, working with Gumshoe.ai, tested this in January 2026. Volunteers ran 12 prompts through ChatGPT, Claude, and Google’s AI a combined 2,961 times. ChatGPT and Google’s AI had less than a 1 in 100 chance of returning the same list of brands in any two runs, and an identical order was rarer still, at less than 1 in 1,000.

The study covered brand and product recommendations, not mortgage brokers, but it is reasonable to expect the same behavior when you ask for brokers. A broker who appears first today may not appear at all tomorrow.

Con 2: You can’t see how the answer was decided

A recommendation is only as good as the reasoning behind it, and AI tools do not show theirs. OpenAI says its own developers cannot fully explain what the model knows or why it sometimes states wrong claims with confidence. When a name lands on a list, you usually cannot tell four things:

• Which sources were weighted, and whether any were paid or promotional.
• How current the information is, since a broker may have changed companies or lost a license.
• What "best" means: lowest rate, fastest closing, most reviews, or best fit for a self-employed borrower. Rate is only one measure, and AI tools cannot judge a broker’s closing record or service.
• Whether the facts are accurate, because OpenAI warns that ChatGPT can sound convincing while giving incorrect information.

Con 3: It leans on third-party sites that have no firsthand experience

When an AI tool searches the web for brokers, it finds what is published about brokers: directory listings, "best of" roundups, review sites, and the brokers’ own marketing pages. Some of those include paid listings or affiliate relationships, and review content can be manipulated. In August 2024 the Federal Trade Commission finalized a rule banning fake reviews and testimonials, including company-controlled review sites that present themselves as independent.

None of these sources can tell you what a broker will do on your loan. This table shows what to confirm and where.

What you need to know	Can AI reliably tell you?	Where to confirm it
Licensing and disciplinary history	No	NMLS Consumer Access
Your actual rate and fees	No	A written Loan Estimate from each lender
Closing track record with your loan type	Not reliably	Ask for recent comparable closed loans, on-time closing history, and references
Which lenders the broker can use	No	Ask the broker to name lenders that fit your file
Past client experience	Not reliably	Reviews across several platforms and direct conversations with past clients
Client service and communication	No	An interview with each broker and conversations with past clients

10 prompts for finding a mortgage broker with ChatGPT in Greater Los Angeles

These work in ChatGPT, Gemini, Claude, Perplexity, and other AI tools. Replace the brackets with your details. Leave out your name, Social Security number, account numbers, and anything else that identifies you.

1. Ask for criteria before names

Start by deciding how you will judge a broker, before any names appear.

2. Make it show its work

A name without a reason is not a recommendation.

3. Separate firsthand from secondhand

This sorts marketing from evidence.

4. Ask about paid placement

The tool may not volunteer this, so ask.

5. Check how current the information is

Brokers change companies, and licenses change status.

6. Ask it to admit uncertainty

A useful answer includes its own weak spots.

7. Get your verification checklist

Do this before you share any documents with anyone.

8. Build your interview questions

Walk into the first call prepared.

9. Test for consistency

Run this in three separate new chats and compare the lists.

10. Compare your own notes

Bring your own research back in, without personal financial details.

How do you turn an AI shortlist into the right Greater Los Angeles mortgage broker?

Follow these five steps and AI becomes a time saver instead of a risk.

1. Confirm the license. Search each name on NMLS Consumer Access, the free public site that shows license status, the states where a person is licensed, employment history, and regulatory actions. Make sure the NMLS number matches what the broker advertises.
2. Read reviews across several platforms. Look for patterns over time instead of a single score, and notice whether the reviews describe loans like yours. Ours are on our reviews page, and the same test applies to them: check them against reviews on sites we do not control.
3. Ask for recent comparable loans. Ask each broker to describe loans closed in Greater Los Angeles in the past year that match your income type, property type, and loan size, and how often files like yours close on the original timeline, without client names.
4. Talk to past clients. Ask for two references with files similar to yours.
5. Interview two or three brokers and compare Loan Estimates from the same day: interest rate, APR, lender fees, and cash to close. According to the CFPB, lenders generally must provide a Loan Estimate within three business days of receiving your application. When you have chosen a broker, the next step is a pre-approval, and our guide explains how mortgage pre-approval works in Los Angeles.

Do not choose on rate alone. A low quoted rate is worth something only if the loan closes on time, and a broker who struggles to close your loan type can cost you more than the rate saves. A track record of closing deals like yours and responsive client service matter as much as price. Ask each broker how many loans like yours they have closed recently, how often those files closed on the original timeline, and who you will talk to when underwriting asks for a new document.

Comparison shopping still pays. A February 2023 Freddie Mac research brief simulated rate shopping on 30-year fixed-rate conventional purchase loans of $250,000 to $350,000 for borrowers with FICO scores of 740 or higher, using daily lender pricing data from 2010 to 2022. In October and November 2022, when lender pricing varied widely, shopping two lenders could have saved as much as $600 a year, and shopping five could have saved more than $6,000 over the life of the loan if the borrower kept it at least five years. Savings depend on how far lender pricing varies on the day you shop, so treat the figures as a guide to scale, not a forecast. Those loan sizes sit below many Greater Los Angeles jumbo balances, so ask each lender to price your actual loan amount. For current market context before you call, see CFG’s Rates page.

Frequently asked questions

Is it a good idea to use ChatGPT to find a mortgage broker?

It is a good way to start and a poor way to finish. ChatGPT and other AI tools can explain loan types, draft questions, and suggest names to research. They cannot verify a license, quote a rate you can lock, or review your documents. Check every name on NMLS Consumer Access and compare written Loan Estimates from at least three lenders.

Why does ChatGPT recommend different mortgage brokers each time I ask?

AI tools generate each answer fresh instead of pulling from a fixed ranking. A January 2026 SparkToro study of brand recommendations found that ChatGPT and Google’s AI had less than a 1 in 100 chance of returning the same list in any two runs. Treat names that repeat as leads and verify each one.

What should I ask ChatGPT or another AI tool when looking for a mortgage broker?

Ask for comparison criteria before names, then ask it to show its sources, separate firsthand information from secondhand, flag paid placement, and admit uncertainty. Use it to write interview questions and a verification checklist, and keep your Social Security number, account numbers, and tax returns out of the chat.

How do I check that a mortgage broker is licensed in California?

Search the broker’s name or NMLS number on NMLS Consumer Access at nmlsconsumeraccess.org. The free public site shows license status, the states where the person is licensed, employment history, and regulatory actions. Confirm that the NMLS number matches the one the broker advertises.

Can I trust the best mortgage broker lists that AI tools cite?

Treat them as leads, not proof. Some directories and roundups include paid listings or affiliate relationships, and review content can be manipulated, which is why the FTC finalized a rule against fake reviews and testimonials in August 2024. Confirm any claim at the source, such as a licensing record, a published ranking, or a written Loan Estimate.

Can ChatGPT or another AI tool tell me today’s mortgage rate?

AI tools can explain what moves rates, but they cannot give you a rate you can lock. Your rate depends on your credit, down payment, property, and loan type, and it changes with the market. A written Loan Estimate from a licensed lender is the document that shows your actual offer.

Is it safe to enter my financial information into ChatGPT or another AI tool?

Avoid entering Social Security numbers, bank or loan account numbers, and full tax returns into any chatbot. Describe your situation in general terms, such as loan size, credit range, and income type, and share documents only with a licensed broker or lender through a secure application.

Is the lowest mortgage rate the best choice when picking a broker?

No. A low quoted rate does not help if the loan does not close on time. A broker’s track record of closing loans like yours and the quality of their client service matter as much as price. Compare rate and fees on written Loan Estimates, then ask each broker about recent comparable closings, typical timelines, and how they communicate during underwriting.

How many mortgage brokers or lenders should I compare?

Compare at least three. A February 2023 Freddie Mac research brief found that shopping two lenders could have saved as much as $600 a year in late 2022, and shopping five could have saved more than $6,000 over the life of the loan. Your savings depend on your loan size and how much lender pricing varies when you shop.

More answers are in the CFG mortgage FAQ.

Choosing a mortgage broker? Start with a real conversation.

Bring your goal, your approximate loan amount, and your income and asset documents, and we will walk you through which loan fits and what it costs. Mark Cohen is ranked the #1 Mortgage Broker in the U.S. in the 2026 Scotsman Guide rankings, and rankings from National Mortgage News (#2) and HousingWire (#5) add consistent industry-wide recognition.

Call 310.777.5401

Sources

• OpenAI weekly active users (more than 900 million, announced February 27, 2026): Search Engine Land, ChatGPT statistics
• OpenAI Help Center, “Does ChatGPT tell the truth?”: help.openai.com
• SparkToro and Gumshoe.ai, “AIs are highly inconsistent when recommending brands or products” (January 27, 2026): sparktoro.com
• Federal Trade Commission, final rule banning fake reviews and testimonials (August 14, 2024): ftc.gov
• CFPB, “Chatbots in consumer finance,” June 2023: consumerfinance.gov
• CFPB, Loan Estimate timing (three business days): consumerfinance.gov
• NMLS Consumer Access: nmlsconsumeraccess.org; explainer from the Conference of State Bank Supervisors: csbs.org
• Freddie Mac, “When Rates Are Higher, Borrowers Who Shop Around Save More,” research brief (February 16, 2023): freddiemac.com

This article is for general information only and is not a loan offer or financial advice. Loan terms, rates, and eligibility vary by borrower and lender.