---
title: "How mortgage platforms win qualified borrowers from AI answers"
description: "By being named, with accurate loan facts, when borrowers ask AI to estimate payments and compare lenders, and by keeping rate claims compliant."
canonical: "https://underneath.agency/resources/mortgage-platforms-leads-ai-search"
published: 2026-10-07
updated: 2026-10-08
publisher: "Underneath (https://underneath.agency/agent)"
entity: "https://underneath.agency/.well-known/entity.json"
---
Guide · AI search

# How do mortgage lenders and marketplaces win qualified borrowers when buyers start with AI?

By being named, and described accurately, when a buyer asks an assistant what they can afford and which lender to use, and by making sure every rate and fee claim an assistant might repeat already follows mortgage advertising rules. A growing share of buyers now run payment estimates and lender comparisons in AI tools before they request a single quote. For a lender or marketplace, that is where a qualified lead is now won or lost.

## The short version

1. AI is now in the mortgage shopping step: in a [Veterans United survey](https://www.veteransunited.com/education/ai-homebuying-survey/) of 859 people in June 2026, 45% of prospective buyers had used AI tools in their home search, up from 37% a year earlier, and 30% used them to shop for the best interest rate.
2. Affordability is the entry point: among prospective buyers who used AI in a [Bank of America study](https://newsroom.bankofamerica.com/content/newsroom/press-releases/2026/06/bofa-study--more-americans-favor-buying-over-renting-for-the-fir.html), 57% used it to estimate affordability, mortgage payments or closing costs.
3. A funded loan barely clears its cost: lenders in the Mortgage Bankers Association’s second-quarter 2026 report, [as reported by HousingWire](https://housingwire.com/articles/imb-mortgage-profits-q2-2026), earned $11,909 per loan and spent $10,936, a pretax profit of $973.
4. Shopping changes what borrowers pay: [Freddie Mac](https://www.freddiemac.com/research/insight/20230216-when-rates-are-higher-borrowers-who-shop-around-save) estimated buyers could save $600 to $1,200 a year by applying with more than one lender, so the shortlist an assistant gives matters.
5. Google shows AI answers on most finance searches: in [our study](https://underneath.agency/research/ai-overviews-frequency-study), financial services and insurance searches triggered an AI Overview 88.0% of the time.

This guide is about how mortgage lenders and marketplaces get found and described in AI answers. It is not mortgage, legal or compliance advice; take rate, fee and referral questions to your counsel and compliance team.

## Who chooses a mortgage platform, and what is a funded loan worth?

A home buyer or homeowner chooses, usually under time pressure; each funded loan carries thin but real margin.

The buyer is a household, not a procurement team. They arrive at a few moments: they need a preapproval letter to make an offer, rates have dropped and a refinance looks worth it, or a life change forces a move. Younger borrowers carry growing weight. A [Cotality survey reported by HousingWire](https://housingwire.com/articles/homebuyers-want-ai-and-human-in-the-loop-cotality-2026-survey) notes that in the US, buyers under 35 account for 37% of originated loans.

The economics explain why lead quality matters more than lead volume. In the MBA’s second-quarter 2026 figures, lenders took in $11,909 in production revenue per loan against $10,936 in cost, for a pretax production profit of $973. The average first mortgage was $386,359, and purchase loans made up 80% of first-mortgage originations by dollar volume among the companies reporting. A lender that pays for leads that never close eats that thin margin quickly. A marketplace earns by passing borrowers to lenders, and its value to those lenders rests on how many of those borrowers fund. Platforms that also offer personal, auto or business loans can compare notes with [how lending platforms reach borrowers through AI](https://underneath.agency/resources/lending-platforms-borrowers-ai-search).

Two kinds of platform compete for the same borrower:

| Platform | How it earns from a borrower | What AI visibility has to deliver |
|---|---|---|
| Direct lender (bank unit, independent mortgage bank, digital lender) | Gain on sale and fees on a funded loan | Applications from borrowers who fit its loan programs |
| Marketplace or comparison platform | Payments from lenders for leads or click-throughs | Borrowers who complete a form and then fund with a participating lender |

## How many borrowers already use AI before they pick a lender?

Between one in five and almost half, depending on the survey; most use it for payment math and research.

- **Veterans United (June 2026).** 45% of prospective buyers had used AI tools. The leading uses were searching for homes (52%) and estimating monthly payments (43%). Shopping for the best interest rate (30%) and comparing lender reviews (28%) followed. ChatGPT was the most used tool (33%), then Gemini (20%). Veterans United is itself a lender, so read this as a lender’s survey.
- **Bank of America (June 2026).** 20% of prospective buyers and current homeowners used AI tools or chatbots for homebuying research in the past year, including 32% of Gen Z. Among AI users, estimating affordability, payments or closing costs led (57%).
- **Cotality (April 2026).** 80% of buyers surveyed assume lenders already use AI. Yet US trust in AI to help find a home fell to 16%, and 64% of buyers worry AI may repeat unverified information instead of relying on validated first-party data.

The surveys differ in who they asked, which explains the spread between 20% and 45%. They agree on the shape: buyers use AI to do the math and narrow options, and they still want people for the high-stakes steps. In Bank of America’s study, 54% preferred human expertise for legal or contractual advice.

The assistants are also becoming storefronts. Zillow launched an [app inside ChatGPT](https://www.housingwire.com/articles/zillow-chatgpt-launch-app-integration/) in October 2025 that guides users from listings back to Zillow, where they can explore financing with Zillow Home Loans. Redfin, now owned by Rocket Companies, [followed with its own ChatGPT app](https://www.housingwire.com/articles/redfin-launched-a-chatgpt-app-to-enable-conversational-home-searches-and-property-exploration-the-move-follows-similar-integrations-by-zillow-and-google-raising-questions-about-mls-data-licensing/). Both route home searchers toward businesses that also finance homes.

## What do borrowers ask AI on the way to a mortgage application?

Payment, eligibility, lender choice and trust questions. The borrower prompts below are our own examples of common questions, not prompts collected from real homebuyers.

| Stage | Illustrative prompt |
|---|---|
| Affordability | “How much house can I afford on $95,000 a year with $400 in car payments?” |
| Program fit | “FHA or conventional with 5% down and a 680 credit score: which costs less over seven years?” |
| Eligibility | “Which lenders do VA loans for a first-time buyer, and what fees should I expect?” |
| Lender choice | “Online lender or local credit union for a first mortgage?” |
| Comparison | “Lender A vs Lender B for a jumbo refinance” |
| Trust | “Is this mortgage company legit? What do borrowers complain about?” |

Each row is a different kind of lead. The affordability question is early and broad; the lender comparison and trust questions arrive close to an application. A reasonable expectation is that the later questions convert at much higher rates, so they deserve the most attention.

## How does an AI answer lead to a funded mortgage?

Through a short chain: the answer shapes the shortlist, and the shortlist decides who gets a quote request. Quotes then decide who funds.

1. **Payment math.** The buyer asks what they can afford. The assistant explains loan types and costs, sometimes citing lender or comparison pages.
2. **Shortlist.** The buyer asks which lenders fit. The assistant names a few, or points to comparison sites.
3. **Trust check.** The buyer asks about reviews, complaints or licensing.
4. **Quote request or form.** The buyer contacts one to three lenders, directly or through a marketplace.
5. **Application, lock and closing.** The lead becomes revenue only here.

Shopping is where the money moves. Freddie Mac found that in 2022, borrowers who applied with two lenders lowered their rate by an average of 20 basis points, and that two rate quotes could have saved as much as $600 a year. A lender left off the shortlist never gets the chance to compete on price. We suggest tracking AI visibility against funded loans and pull-through, not just form fills, and asking applicants where they first heard of you; analytics often miss AI-assisted visits, as explained in [why analytics miss AI visibility](https://underneath.agency/resources/why-analytics-miss-ai-visibility).

## What decides whether an assistant names a lender or marketplace?

In finance, sources that already rank, comparison sites and review platforms weigh heavily; platforms document little about choosing lenders.

**Documented by the platform.** A question such as “best FHA lender for first-time buyers” may, by Google’s own account, set off [“query fan-out”](https://developers.google.com/search/docs/appearance/ai-features) in AI Overviews and AI Mode: several related searches run before the answer is written. Because a mortgage can affect a household’s financial stability, Google’s guidance holds such topics to a higher bar, and its systems [weigh experience, expertise, authoritativeness and trust more heavily](https://developers.google.com/search/docs/fundamentals/creating-helpful-content) there. Neither Google nor OpenAI publishes how it picks one lender over another.

**Observed in our studies.**

- **Ranking pages feed mortgage answers.** Of the AI Overview citations for finance and insurance searches in [our citation study](https://underneath.agency/research/ai-overview-citations-study), 34.8% were page-one Google results; nerdwallet.com alone showed up in 10.0% of every AI Overview we sampled.
- **Comparison sites carry over into answers.** In [our hidden-searches study](https://underneath.agency/research/ai-hidden-searches-study), when ChatGPT’s own search named a source such as NerdWallet, the answer cited it 44.0% of the time, against 8.1% when it did not.
- **Trust questions go to review platforms.** In [our reputation study](https://underneath.agency/research/is-it-legit-ai-reputation-study), 88.0% of answers to “is this brand legit?” cited a review or complaint platform.
- **Video explains the math.** In [our YouTube study](https://underneath.agency/research/ai-overview-youtube-videos-study), AI Overviews for financial services searches cited a YouTube video on 49.0% of searches, although Google showed one on page one for only 8.0%.

**Our inference.** For a lender, the facts that decide whether an assistant can recommend you are concrete: which loan programs you offer, in which states you are licensed, minimum credit and down payment requirements, how fast you close, and what borrowers say after closing. A lender whose program pages state these plainly gives an assistant something to repeat. One whose details sit behind a lead form gives it nothing, and the assistant may name a comparison site or a rival instead. Insurers face a similar state-by-state test, covered in [how insurance carriers win quotes from AI](https://underneath.agency/resources/insurance-companies-customers-ai-search).

## How do RESPA and TILA rules shape mortgage GEO?

They apply to your own content and referral arrangements; they make accuracy and neutral presentation the safest route to visibility.

**Advertising under Regulation Z.** The Truth in Lending Act’s [Regulation Z](https://www.ecfr.gov/api/renderer/v1/content/enhanced/current/title-12?part=1026&section=1026.24) says that if an advertisement states a rate, it must state it as an “annual percentage rate,” using that term. It lists “triggering terms,” such as a payment amount, that require further disclosures, and it bars misleading comparisons using rates or payments that apply for less than the full loan term. Rate tables, affordability calculators and “as low as” pages are exactly what an assistant may quote. We infer that a page written to these rules is also the page least likely to be misquoted.

**Referrals under RESPA section 8.** In a [2023 advisory opinion](https://www.govinfo.gov/content/pkg/FR-2023-02-13/pdf/2023-02910.pdf), the CFPB said a digital mortgage comparison-shopping platform violates section 8 if it gives enhanced placement or steers consumers to participants based on the payments it receives, rather than on neutral criteria. Federal guidance has since shifted: in [May 2025 the CFPB withdrew](https://www.govinfo.gov/content/pkg/FR-2025-05-12/html/2025-08286.htm) a long list of guidance documents, including its 2024 circular on preferencing and steering by digital intermediaries, and said the withdrawal “is not necessarily final.” The statute itself did not change, and state regulators enforce their own rules. Check the current status of each document with counsel.

For a marketplace, the practical point is that “best lender” rankings and comparison tables are both the content assistants like to cite and the content regulators have examined. Publishing the criteria behind any ranking, and keeping paid placement clearly separate, protects you on both fronts. For how ranked lists feed AI recommendations, see [which pages to target to show up in AI recommendations](https://underneath.agency/resources/best-of-lists-ai-recommendations).

## What does GEO work look like for a mortgage platform?

Generative engine optimization (GEO) makes your loan programs, costs and reputation easy for assistants to find and repeat accurately.

For a lender or marketplace, the work usually covers:

1. **Program pages in plain text.** One page per loan program (FHA, VA, conventional, jumbo, refinance) stating who qualifies, licensed states, typical timelines and documents needed, reviewed by compliance.
2. **Compliant rate and cost content.** APR shown wherever a rate appears, triggering-term disclosures in place, and a date on every rate example, so a quoted figure is not stale or misleading.
3. **Affordability explainers.** Clear, worked explanations of payments, closing costs and mortgage insurance, the questions buyers bring to AI first. Short explainer videos with accurate captions belong here too.
4. **Comparison and editorial coverage.** Accurate listings and reviews on the finance comparison and editorial sites assistants search and cite.
5. **Reputation work.** Monitor review and complaint platforms, answer complaints, and fix their causes; assistants summarize what those platforms say.
6. **Entity facts.** Consistent company name, licensing details, ownership and contact information across your site and directories, so assistants do not confuse you with a similarly named lender.
7. **Search foundations.** Because finance answers lean on page-one results, keep program and explainer pages ranking well in Google and Bing.

If an assistant misstates your loan limits, licensed states or program rules, [our guide to fixing wrong brand information in AI answers](https://underneath.agency/resources/fix-wrong-brand-information-in-ai-answers) shows how to correct it. For marketplaces with two-sided dynamics, [how an online marketplace wins buyers and sellers when people ask AI](https://underneath.agency/resources/marketplace-buyers-sellers-ai-search) covers the supply side. No one can guarantee that an assistant will name a lender; GEO makes the evidence it finds accurate, current and verifiable.

## Where is the mortgage evidence still thin?

On conversion: surveys show borrowers use AI, but nothing public links AI visibility to applications or funded loans.

- **Surveys come from interested parties.** Veterans United and Bank of America are lenders, and Cotality sells data to the industry. Their samples and questions differ, which is why the AI-use figures range from 20% to 45%.
- **No public conversion data.** We found no public figures connecting AI answers to mortgage applications, lock rates or funded volume. The cross-industry evidence, thin as it is, is reviewed in [does AI visibility drive business results?](https://underneath.agency/resources/does-ai-visibility-drive-business-results).
- **Rules are in motion.** Federal guidance on comparison platforms changed in 2025 and may change again; how regulators view AI-generated rate quotes has not been tested in public.
- **Our studies are snapshots.** Our citation and frequency findings come from US searches in September 2026; answers change between assistants and over time.

## Where should a mortgage lender or marketplace start?

Start by asking assistants the questions your borrowers ask, then check every rate, program and licensing fact in the answers.

A useful first review covers affordability and program questions in your main states, lender-choice and comparison questions, and the reviews and “is it legit” questions about your company. It shows whether you are named, which lenders and comparison sites appear instead, and whether your programs, licensing and costs are described correctly.

If your growth depends on borrowers who apply and fund, [talk to us about a review of your mortgage platform in AI answers](https://underneath.agency/contact). We will map where assistants send borrowers in your markets, list the facts they get wrong or cannot find, and plan the program pages, coverage and reputation work, reviewed with your compliance team, that give you a fair chance at the shortlist. Lenders and marketplaces can see on our [generative engine optimization service](https://underneath.agency/services/generative-engine-optimization) page how that program and reputation work is run, with compliance review built into each step.

## Frequently asked questions

### Do home buyers really use ChatGPT to choose a mortgage lender?

Some do. In Veterans United’s June 2026 survey, 30% of prospective buyers used AI to shop for rates and 28% to compare lender reviews. Most AI use is still payment math and research.

### Can an AI assistant quote our mortgage rates?

It can repeat rates from pages it finds, including old ones. Date your rate examples, show the APR wherever a rate appears, and keep historical rate pages clearly labeled.

### Do mortgage comparison sites help or hurt a lender in AI answers?

Often they help. Finance answers cite comparison sites frequently, and in our hidden-searches study ChatGPT cited a named source 44.0% of the time. Accurate listings there matter.

### Does GEO replace paid mortgage leads?

No. It is a separate channel. Paid leads deliver contacts now; GEO improves how assistants describe you over time. Measure both against funded loans, not form fills.

## Sources

- Veterans United Home Loans (2026-07-02), [New Survey: More Homebuyers Turning to AI Tools in 2026](https://www.veteransunited.com/education/ai-homebuying-survey/)
- Bank of America (2026-06-23), [BofA Study: More Americans Favor Buying Over Renting for the First Time Since 2023](https://newsroom.bankofamerica.com/content/newsroom/press-releases/2026/06/bofa-study--more-americans-favor-buying-over-renting-for-the-fir.html)
- HousingWire (2026-04), [Homebuyers want AI and human in the loop: Cotality 2026 survey](https://housingwire.com/articles/homebuyers-want-ai-and-human-in-the-loop-cotality-2026-survey)
- HousingWire (2026-08), [IMB mortgage profits, Q2 2026](https://housingwire.com/articles/imb-mortgage-profits-q2-2026)
- Freddie Mac (2023-02-16), [When Rates Are Higher, Borrowers Who Shop Around Save More](https://www.freddiemac.com/research/insight/20230216-when-rates-are-higher-borrowers-who-shop-around-save)
- HousingWire (2025-10-06), [Zillow, ChatGPT launch app integration](https://www.housingwire.com/articles/zillow-chatgpt-launch-app-integration/)
- HousingWire (2026), [Redfin rolls out ChatGPT app for real estate searches](https://www.housingwire.com/articles/redfin-launched-a-chatgpt-app-to-enable-conversational-home-searches-and-property-exploration-the-move-follows-similar-integrations-by-zillow-and-google-raising-questions-about-mls-data-licensing/)
- eCFR (2026), [12 CFR 1026.24, Advertising (Regulation Z)](https://www.ecfr.gov/api/renderer/v1/content/enhanced/current/title-12?part=1026&section=1026.24)
- Federal Register, CFPB (2023-02-13), [Digital Mortgage Comparison-Shopping Platforms and Related Payments to Operators](https://www.govinfo.gov/content/pkg/FR-2023-02-13/pdf/2023-02910.pdf)
- Federal Register, CFPB (2025-05-12), [Interpretive Rules, Policy Statements, and Advisory Opinions; Withdrawal](https://www.govinfo.gov/content/pkg/FR-2025-05-12/html/2025-08286.htm)
- Google Search Central (2025), [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features)
- Google Search Central (2025), [Creating helpful, reliable, people-first content](https://developers.google.com/search/docs/fundamentals/creating-helpful-content)
- Underneath (2026), [When does Google show an AI Overview?](https://underneath.agency/research/ai-overviews-frequency-study)
- Underneath (2026), [Do AI Overviews cite the pages that rank?](https://underneath.agency/research/ai-overview-citations-study)
- Underneath (2026), [The hidden searches AI assistants run before they answer](https://underneath.agency/research/ai-hidden-searches-study)
- Underneath (2026), [“Is this brand legit?” How AI assistants build a reputation](https://underneath.agency/research/is-it-legit-ai-reputation-study)
- Underneath (2026), [The YouTube videos Google’s AI cites are small](https://underneath.agency/research/ai-overview-youtube-videos-study)

---

This is the Markdown twin of https://underneath.agency/resources/mortgage-platforms-leads-ai-search. The HTML page is canonical. Publisher: Underneath, https://underneath.agency/agent. Site index: https://underneath.agency/llms.txt.