We were curious about a simple question that more and more borrowers are now asking out loud: when you open ChatGPT and ask for a mortgage broker, who does it actually name? So we ran the experiment. Across the audits and tests we run, we have prompted ChatGPT and other answer engines for mortgage brokers hundreds of times, varying the wording, the city, and the borrower’s situation. The pattern is consistent enough to be useful, and it explains why so many capable brokers are invisible in AI search while a few names show up again and again.
The short version: a ChatGPT mortgage broker recommendation is not random and it is not a popularity contest in the way Google rankings can feel. It is a trust-and-verifiability contest. ChatGPT recommends the brokers it can read clearly, confirm from more than one source, and quote without guessing. If your business is hard to verify, it gets left out, no matter how good you are at the actual work.
How we ran the test
To keep our observations honest, we used a few rules. We asked in plain language, the way a real borrower would. We tested generic prompts (“find me a good mortgage broker”) and local, intent-rich prompts (“who is the best mortgage broker in [city] for first-time buyers”). We repeated each prompt across fresh sessions to see what held steady versus what was noise. And we noted not just the names ChatGPT gave, but the sources it pointed to when it cited anything.
What we care about here is repeatability. A name that appears once might be a fluke. A name that appears across many sessions, across slightly different wording, is a signal that the model has settled on that business as a safe answer.
Generic prompts versus local prompts
The single biggest factor in whether ChatGPT names a specific broker is how specific the question is. Generic prompts almost never produce a local name. Instead, ChatGPT returns process advice (“compare rates, check reviews, ask about fees”) and points toward large national lenders and aggregator brands it is highly confident exist. The moment a prompt includes a city and an intent, the answer changes. That is when individual brokers start appearing by name.
Who ChatGPT actually recommends
When the prompt is local and specific, the recommendations fall into a few predictable buckets. We are not naming individual brokers here because the specifics shift by market and over time, but the categories are stable.
| Prompt type | What ChatGPT tends to recommend | Why |
|---|---|---|
| Generic ("find me a mortgage broker") | National lenders, big brands, process advice | Low risk; the model is confident these exist and stay relevant |
| Local + intent ("best broker in [city] for first-time buyers") | Named local brokers with strong digital footprints | The model can verify them across reviews, profiles, and their site |
| Niche ("broker for self-employed buyers in [city]") | Brokers who publish content addressing that exact niche | Answer-first pages match the question word for word |
The brokers who win the local and niche prompts share a profile. They are not always the biggest shop in town. They are the ones who are easy to verify and easy to quote. That distinction is the whole story of AI recommendations.
The sources ChatGPT leans on
When ChatGPT cites anything for a mortgage recommendation, the citations cluster into a short list of source types. In our testing, these are the ones that show up most often:
- Google Business Profile and Maps listings. A complete, active profile is the most consistent signal. It confirms the broker is real, local, and operating.
- Review platforms. Steady, recent reviews give the model both a confidence boost and language to quote (“clients praise their communication”).
- Established directories and association pages. Listings in lender networks, NMLS-linked pages, and trade associations act as corroborating evidence.
- Local news, community, or partner mentions. Even a single credible local citation can tip a broker from invisible to named.
- The broker’s own website, when it is well structured. Answer-first pages and clean schema let the model read the business directly rather than guessing.
The key pattern: brokers cited across more than one of these surfaces are far more likely to be named in the answer itself, not just buried in a source list. One strong source helps. Several agreeing sources is what gets you recommended. If you want the deeper mechanics of which directories and platforms carry the most weight, our breakdown of the most-cited sources when you ask AI for a realtor covers the same dynamics in an adjacent industry.
Why the same few brokers keep winning
One thing surprised us early and then stopped surprising us: within a given market, ChatGPT tends to recommend the same small set of brokers across many sessions. This is not favoritism. It is the model doing the rational thing. Once a broker has a consistent, multi-source footprint, recommending that broker is the lowest-risk choice the model can make. It has the evidence. It can quote the reviews. It can confirm the contact details. So it returns to that name.
The flip side is harsh for everyone else. A skilled broker with a thin or inconsistent digital footprint simply does not enter the consideration set. The model is not weighing them and deciding against them; it cannot see them clearly enough to risk naming them. Consistency of your name, address, and phone across the web matters more than most brokers expect, which is exactly what we found in our study on whether NAP consistency affects AI recommendations.
What this means if you are a mortgage broker
The encouraging finding from all our testing is that the recommendation set is changeable. It is not locked. Because it rewards clarity and consistency rather than raw size, a focused local broker can break in faster than a national brand can defend a city-level prompt. We have watched it happen. Keith Akada, a Seattle mortgage broker, went from invisible in AI search to the number-one AI-recommended broker in his market in about six weeks, generating roughly 30 leads and four closed deals in that window, simply by becoming the easiest broker in his city for AI to verify and quote.
If you want to move into the set ChatGPT recommends, here is the priority order we use, drawn from what actually changes outcomes:
- Complete and maintain your Google Business Profile. Every field filled, category correct, hours current, and kept active. This is the single highest-leverage move.
- Fix your NAP consistency. Make your name, address, and phone identical everywhere they appear. Inconsistency makes the model uncertain, and uncertainty means it picks someone else.
- Build a steady review cadence. Recency and volume both matter. Reviews give the model both confidence and quotable language.
- Publish answer-first content. Write pages that answer the exact questions borrowers ask, with the answer up top. This is how you win the niche prompts.
- Add structured data and an llms.txt file. Organization and FAQ schema, plus an llms.txt that tells AI engines what you do and where, helps them read you accurately rather than inferring.
The mistake most brokers make
The most common error we see is treating AI visibility like traditional SEO, where the goal is to outrank competitors on a results page. AI recommendation is different. The goal is to become un-skippable: the name the model can verify so easily that leaving you out would be the harder choice. That shift in framing changes what you work on. You stop chasing keywords and start building a footprint that any answer engine can confirm from several directions at once.
The bottom line
When you ask ChatGPT for a mortgage broker, it recommends the businesses it can trust and quote, and it keeps recommending them because consistency compounds. The names that win are not always the biggest; they are the clearest. That is genuinely good news if you have been doing strong work but staying quiet about it online. The path into the recommendation set is concrete, it is faster than most brokers assume, and it is built from the same fundamentals every time: be verifiable, be quotable, and be the answer to the exact question a borrower is asking.