We asked ChatGPT who to hire in Park City, Utah, live during a session for luxury agents. Here is who it named, why, and the five signals AI checks before it recommends an agent.
By Andrew Hong. From Session 01 of Become the Agent AI Recommends, presented live 22 September 2026.
The short answer: ChatGPT recommends the agent whose specialty it can read and verify. When we asked it live who to hire in Park City, Utah, nearly every agent it named came with a specialty attached, and only a few came with a sales figure. Agents who never state a specialty anywhere AI can check get skipped.
This post comes from Session 01 of Become the Agent AI Recommends, a free four-part live series for luxury agents. I'm Andrew Hong. I'm not a real estate agent. I'm the person behind a lot of them. Tobe Agency has managed more than 100 agent and brokerage websites since 2016, and over the last 18 months we've run more than 50 AI visibility audits for luxury agents. You can download the full slide deck free, no email needed.
A referral used to end the search. Now it starts one.
Referrals still drive the business. In NAR's 2025 Profile of Home Buyers and Sellers, 43% of buyers found their agent through a referral, and 81% of sellers contacted only one agent before hiring. What changed is the step in between. A friend gives your name, and before anyone picks up the phone, they ask AI about you:
That step is already mainstream. Realtor.com surveyed 1,000 Americans who were buying or selling a home, or had in the past two years, and 82% said they use AI for housing market information, led by ChatGPT (67%) and Gemini (54%) (October 2025). A few more numbers explain why:
When we pull keyword data for "best luxury real estate agent in [city]," Google search volume is close to zero. Those questions haven't gone away. They've moved into AI.
So the real question is what AI knows about you. If it can't confirm what the referral said, the referral gets weaker. If it names someone else, you may lose the one call.
SEO (search engine optimization) is how you rank on Google and Bing when someone searches for an agent in your market: the local map pack, the organic results, your Google Business Profile.
AEO (answer engine optimization) is making sure an answer engine can find you, verify you, and recommend you. For an agent, that means content that answers the exact questions buyers and sellers ask, clearly enough that Google lifts it into an AI Overview or a featured answer. FAQ pages and schema markup do most of this work. Many agent websites are more than five years old and were never built with schema, so they start at a disadvantage for these answers.
GEO (generative engine optimization) is earning a place inside the generated answer itself: whether ChatGPT, Gemini, and Perplexity understand who you are, and whether they name you when someone asks, "Who should I use to buy in Park City?"
Good SEO is still the foundation for the other two. If you only do SEO, you're solving a third of the problem.
GEO is where I see the best leads. The person asking has already given their AI assistant a lot of context: their budget, their family, how they spend winters. By the time they reach you, the matching is done, and you spend less time qualifying and more time on deals that close.
During the session I ran three questions live in ChatGPT, logged out, in a fresh incognito window with no history:
Here is what came back.
It looked like a local search result. ChatGPT opened with a map and star ratings, much like a Google Business Profile listing, then a short list of agents. Ratings sat front and center.
Every name came with a reason. The reasons were specific: a community (Deer Valley, Empire Pass, Promontory), a price band ($3 million and up), off-market access. A few answers mentioned sales volume. Nearly every one stated a specialty. Agents with half a billion dollars in annual volume sat on the same list as individual agents with a sharper focus.
The seller question went to the clearest brand. One team had built its entire identity around Deer Valley: the team name, the website, the neighborhood pages. ChatGPT named them for the ski-in ski-out question and described decades of Deer Valley focus. Another brokerage on the list was named after ski resort property. Both are signals AI can read without guessing.
The referral check pulled their history. When I asked about that Deer Valley team by name, ChatGPT listed the neighborhoods they work (Empire Pass, Deer Crest, Silver Lake), their tenure, their reviews, and their brokerage affiliation. The brokerage never came up in the first two answers. For the specific question, the team's own brand carried more weight than the franchise behind it.
Both answers ended with questions to ask the agent. Agents I work with tell me they can spot a prospect who has talked to ChatGPT, because the prospect asks exactly those questions. Run the test on your own market and you'll know them in advance.
One more thing from the referral check. Many agents tell me they don't like to talk about their track record. They'd rather let the work speak. If you've worked one community for 20 or 30 years, that history is the proof AI is looking for. When you leave it unsaid, AI describes your competitor's history instead.
The weekend before this session, my basement flooded. Utah has dozens of restoration companies, and Google Maps gave me 15 pins with no way to tell them apart. So I asked Gemini. It named three. I checked their Google profiles: strong reviews, online booking. The first company I called answered with an AI receptionist, and nobody showed up. The second answered with a person, had someone at my house in under 24 hours, and is writing the bid now. I'm not getting a third quote. I don't have the time.
Your buyers and sellers are doing the same thing with agents. Two or three names, then one call.
AI's job is to match specialists to situations, and it can only match a specialist it can verify. It does that by cross-checking your site against independent sources, the same trust stack it uses for every business.
Picture a table with a "best for" column. If AI can't fill in that cell for you, you don't make the table. "Proactive," "hardworking," and "20 years in the market" don't fill it. "Ski-in ski-out in Deer Valley," "Bay Area relocations to Park City," and "waterfront listings over $5 million" do.
Buyer and seller questions get different answers, too. You need a clear signal for each side you want to win.
Niching down feels risky to a lot of agents. But if you've never written your specialty on your website, your Google Business Profile, or your Zillow profile, AI has nothing to match you to.
This summer we audited a luxury team on Florida's Gulf Coast (anonymized here):
Their pre-2014 sales history lived on a website from the late 1990s, branded under a brokerage that no longer exists. Key pages were set to "noindex," which tells search engines not to read them. AI could not connect four decades of proof to the team.
Forty-one years of proof, and none of it where AI looks.
1. Credentials it can verify. Mention your ILHM membership and your CLHMS or Guild designation, and keep your Institute profile current. AI knows these organizations and can check them. Press, awards, and published production carry weight because they live on someone else's site. Local press counts, often more than a national mention. FastExpert, where agents submit production for verification, is a source AI reads. Credentials support a specialty. They don't replace one.
2. One consistent identity. Your name, address, phone number, and brokerage need to match everywhere, down to the letter. "Suite 200" and "#200" can read as two different agents. If an old brokerage still has a page up for you, ask them to remove it or set it to noindex. Start with the heavy hitters: Google Business Profile, Zillow, Realtor.com, and Homes.com. Then FastExpert, Bing Places, Apple Maps, LinkedIn, and your ILHM profile. Check the old Yelp and Better Business Bureau listings you forgot about, too. (The full NAP and citation cleanup.)
3. Content it can cite. Your buyer and seller pages should name the communities you serve, explain your process, and say what you do differently when you market a listing. Add neighborhood pages for the areas you want to own, and buyer and seller FAQs in plain language. Publish your market report as a web page. A PDF is much harder for AI to read and cite.
4. Reviews that back you up. Across our audits, 30 or more Google reviews is roughly where AI starts naming agents in competitive markets. Aim for about a third from sellers if you want listings, and ask sellers to mention that you listed their home and how it sold. Zillow and Realtor.com reviews count too. Agents who run a real review push usually see more calls not long after. (Why reviews decide the map pack and the AI answer.)
5. A website AI can read. Own your domain and your content, not only a brokerage template. Keep your bio, markets, and recent sales current. Use schema markup so AI can read who you are, and make sure nothing important is set to noindex. If your site hasn't been rebuilt in five to ten years, it's worth an audit. (What an AI-search website build includes.)
Free download, no email needed
Get the Session 01 slides
The full deck from the session: the live Park City test, the five signals AI checks, and what to fix this week and this month. Keep it, and share it with your team.
Score yourself on all five signals. The free AI Visibility Checklist has 25 points and takes about 10 minutes.
Want the full picture? The AI & Search Intelligence Blueprint is $999, one time. It shows what ChatGPT, Gemini, Claude, Perplexity, and Google say about you, with exact quotes; your share of answers against your top three competitors; a 13+ directory audit; and a six-month prioritized roadmap. It's credited in full toward any implementation program started within 14 days.
Next session: The Brand AI Can Repeat. Tuesday, October 27, 3 PM MT / 5 PM ET. Your story, your message, and your market in one sentence buyers and AI can repeat. Register on Luma.
It matches a specialty to the situation in the question, then checks that specialty against sources it can verify: your website, Google Business Profile, Zillow, Realtor.com, directories, reviews, and press. In our live Park City test, nearly every agent it named came with a stated specialty, and only a few came with a sales figure.
SEO is how you rank in Google and Bing results. AEO is whether your content answers buyer and seller questions clearly enough to appear in AI Overviews and featured answers. GEO is whether ChatGPT, Gemini, and Perplexity understand who you are and name you when someone asks who to hire.
State a specialty, keep your name, address, phone, and brokerage identical on every profile, publish neighborhood pages and plain-language FAQs, build toward 30 or more Google reviews that include sellers, and keep a website you own that AI can read. Start by asking ChatGPT and Gemini the buyer and seller questions for your market, logged out, to see where you stand.
Because the person you were referred to often asks AI about you before they call. If your profiles disagree, your history lives on an old site, or no source states your specialty, AI can't confirm what your referral said, and it may suggest an agent it can confirm.
Some. Production published on a third-party site helps AI verify you, and it matters more on listing questions at specific price points. In our test, production alone didn't earn a spot. A clear specialty did.
There's no fixed number. Across our audits, 30 or more Google reviews is roughly where AI starts naming agents in competitive markets. Recent reviews from both buyers and sellers count most.
This post comes from Become the Agent AI Recommends, a free four-part live series from Tobe Agency. Remaining sessions are October 27 (The Brand AI Can Repeat), November 17 (The AI-Ready Website), and December 15 (The Google Business Profile Advantage). Each one is recorded, and the replay goes to everyone who registers.
About the author. Andrew Hong founded Tobe Agency in 2016. He has trained thousands of real estate agents through ILHM, Colibri Real Estate, and REALM, manages more than 100 agent and brokerage websites, and has run more than 50 AI visibility audits for luxury agents. He is not a real estate agent. He's the person behind a lot of them.