How AI Builds the "Top Real Estate Agent in Your City" Shortlist

Type it yourself, right now, about your own city. Top real estate agent in Dallas. Best realtor in Naperville. Whatever your market is.

You get three names. Sometimes two, sometimes five, never forty. A short paragraph on each, a star count, maybe a sentence about what they are known for.

That is not a search result anymore. That is a shortlist, and somebody built it in under two seconds without asking anyone in your market for permission.

Nobody is ranking agents. Something is assembling an answer.

Here is the thing most agents get wrong about this. You picture a leaderboard, with the best agent at the top and everyone else beneath in order of merit. There is no leaderboard.

An engine is doing something much simpler and much more brutal. It is trying to write two sentences about a local professional that it will not be embarrassed by. To do that it needs facts it can stand behind: who you are, where you operate, that the phone number is right, that other people said specific things about working with you, that all of this agrees with itself everywhere it appears.

Machines don't recommend the best agent. They recommend the one they understand. The shortlist is not a ranking of quality. It is a list of the agents the machine could describe with confidence.

The four things it is looking for

When you break down who keeps showing up in these answers, the same four ingredients are there every time. None of them are clever.

• One consistent identity. Name, address, phone, the same in every place a machine can find them. Roughly 74% of businesses get excluded from local results over mismatched name, address and phone data. Excluded, not demoted.
• A profile with weight behind it. Google Business Profile signals are about 32% of local ranking and reviews about another 20%. That is over half the picture sitting in two places most agents treat as a set-it-and-forget-it chore.
• Reviews that contain sentences, not adjectives. "Great agent, highly recommend" is unquotable. "It had sat four months, she repriced it, under contract in nine days" is an answer the engine can lift and use.
• A place of your own that explains your market. Pages about the city, the neighborhoods, the price bands you actually work in, written like a person who lives there. Your listing feed is not that. Your feed describes houses; you need something that describes you.

Why the same three names keep winning

Once an engine has enough clean material on someone, that agent becomes the cheap, safe answer. It costs the machine nothing to name them again. So they get named again, and each mention becomes another citation of the same record, and the record gets even easier to trust.

This is the compounding you are up against. Not a bigger ad budget. A head start on being legible.

Evan Downey is one solo agent at eXp in Dallas, a market with ten to fifteen thousand agents in it. He is the name the machines hand out for Best Realtor Dallas, and roughly $250,000 in commission last year came from Google, ChatGPT and Grok. He did not outspend the big teams. His record was simply easier to read than theirs.

Ryan Comstock is the same lesson from the other direction. Twenty years, 250 Google reviews, and he was on page three while agents with about fifty reviews sat above him. His name, address and phone did not match across the web. As he put it, he was confusing Google. Once one clean record existed, twenty sales and about $250,000 followed in the first year, and buyers started arriving from Copilot and Claude.

Where the portals fit, and where they do not

A fair question at this point: I pay Zillow, doesn't that cover me. It does not, and not because the portals are bad at what they do. They are excellent at it. They collect the demand and rent it back to you.

But when an engine builds a shortlist of agents, it is looking for an agent it can describe, not a lead form it can fill. Mandy Wilson cancelled a $60,000 a year Zillow spend and took seven listings in her first eight months once her own record was doing the work. Same market, same agent. The difference is that the demand arrived with her name on it instead of being sold to her by the click.

Take your own numbers on that one. Whatever you paid for leads last year, divide it by the closings that actually came from it, and ask what that same money would buy if it went into an asset you keep.

You are not the problem

If you are not on your city's shortlist, the reason is almost never effort. Most agents I look at have worked harder than the agent who is on it.

What happened is that you were sold your presence in pieces. A website from one vendor, a review request tool from another, a Google profile a brokerage assistant set up in 2019 with the old office suite number, a couple of directory listings you never claimed. Every piece was fine. Nothing connected. So the machine reading all of it could not decide which version of you was true, and it did what it always does with uncertainty. It left you out and named someone simpler.

Nothing about you needs fixing. The record does.

How to see your own shortlist honestly

Five minutes, and it is uncomfortable in a useful way.

• Ask three engines the same question a seller would ask. ChatGPT, Gemini and Google's AI answer: top real estate agent in your city. Write down every name that comes back.
• Ask a follow-up the way a real seller does. Who is best for a $700,000 listing in this neighborhood, or who handles relocations here. Watch whether the names change. That second question is where most local shortlists are actually won.
• Look up the agents that were named. Not their production. Their profile, their review text, their site. You will usually find someone with fewer years and cleaner information than you.
• Then check your own name, address and phone in four places: your Google profile, your website footer, your brokerage page, one old directory. Count the differences. That number is your homework.

The honest constraints

This is not a switch you flip. Expect six to twelve months of consistent work before engines reliably hand out your name, and an agent who would quit at month four should not start. It also asks something small of you every week rather than something big once: one review ask, one piece of local content, answering setup questions when they come.

And partly yes, this is what we do for a living. We keep the record straight, the profile fed and the market pages real, and our guarantee is first page among the top 20 agents in a market within 12 months or we keep working at our own cost.

But you can start the diagnosis today without us, and you should, because next spring a seller in your market will sit down on a Tuesday night, type four words, get three names, and decide before anyone's phone rings.

Later is not a decision. Sellers call you. Not you calling sellers.

Frequently asked questions

How does AI decide which real estate agents to include in a top agent list?

It assembles the answer from information it can verify and describe: a consistent name, address and phone across the web, an active and well-reviewed Google Business Profile, review text with specific detail in it, and pages that explain the agent's market. Confidence, not merit, decides inclusion.

Do more reviews put me on the shortlist?

Not by themselves. Reviews are roughly 20% of local ranking and they matter, but an agent with 250 reviews and mismatched contact data can still sit on page three behind agents with 50 reviews and one clean record.

Can a solo agent beat a large team in AI results?

Yes, and it happens regularly. Legibility is not bought at scale. A one-person business can hold a cleaner, more consistent record than a forty-person team, and the engine prefers the record it can read.

How long does it take to start appearing in these answers?

Plan on six to twelve months of consistent work: one clean identity, steady reviews with real detail, and genuine local content. Early signs usually show sooner in longer, more specific questions than in the broad top agent query.

Related reading: What you actually pay per closing for portal leads