Ask ChatGPT, Perplexity, or Google’s AI the question buyers and sellers now type every day: “Who is the best real estate agent in my city?” You get a confident, organized answer. Names, brokerages, reasons. In most markets, that answer is wrong.
We tested it across 50 top U.S. real estate markets. We compared the agents AI recommends against RealTrends, the verified production rankings the industry treats as the standard of record and the Wall Street Journal publishes every year. RealTrends ranks agents on audited transaction sides and sales volume. Not reputation. Not marketing. Closed business.
The gap was not close.
In 39 of 50 markets (78%), the single highest-producing agent in town did not appear in AI’s recommendation at all. In 31 of 50 markets (62%), AI named none of the top five producers.
What AI Recommends Instead
When AI skips the real top producers, it does not return nothing. It returns someone. The replacements fall into a few predictable buckets.
When asked for the best agent in Los Angeles, AI returned a roster of television personalities, the stars of Million Dollar Listing and Buying Beverly Hills. The verified number one in Los Angeles by volume is Cindy Ambuehl at over $315 million. She was not on the list. Neither was the verified number two, who closed 142 transaction sides in the annual rankings.
In other markets AI surfaced whoever the review platforms ranked highest, rather than the most closed business. In several markets the top AI-recommended “best agent” traced back to a paid press release on a newswire service. In many, AI simply repeated whichever agent had published “the number one realtor in this city” on their own website.
None of these signals measure whether an agent actually sells homes. AI is not lying. It is doing exactly what it was built to do.
Why This Happens
AI cannot see production data. It has no access to RealTrends, to the MLS, or to a brokerage’s internal numbers. When you ask it for the best agent in a city, it assembles an answer from the structured, third-party information about that market that already exists on the open web, and it weights whatever is repeated most confidently across independent sources.
So AI does not measure who is best. It measures who the third-party web says is best. Where a market’s online footprint is built from television fame, AI returns celebrities. Where it is built from review counts, AI returns the most reviewed. Where a top producer has quietly closed $300 million in homes but has no structured third-party presence, AI cannot see them, and neither can the consumer asking for a referral.
This is also why an agent’s own website does not fix the problem. An AirOps analysis of 21,311 brand mentions across ChatGPT, Claude, and Perplexity found that 85% came from external third-party domains and only 13.2% from the brand’s own website. An agent declaring themselves the best on their own About page is the one signal these systems are built to discount.
The Proof Is in Where AI Got It Right
The 11 markets where AI named the real number one are the most important finding in this study, because they show the mechanism is real and fixable.
In every one, the same thing was true: structured, verified, third-party data about that agent existed somewhere AI could read it. In Miami, AI surfaced Dina Goldentayer, the verified number one. In Tampa, Jennifer Zales. In Minneapolis, Artemisa Boston. In Washington, DC, AI correctly named the verified number one and an agent from the verified top five, because a news feature and that agent’s own structured “number one in DC” positioning had pushed the data onto the web AI reads.
That last case is the trap. It tempts agents and brokerages to think the answer is to issue a press release in every market. It is not. A press release is a one-time, unstructured signal that AI mostly ignores and occasionally rewards by accident, and it decays the moment the news cycle moves on. You cannot build a referral channel on luck, and you cannot do it across hundreds of markets and thousands of agents by hand.
There is a deeper reason a brokerage cannot solve this with its own announcements. The entire reason AI discounts agent websites is that it down-weights what a business says about itself. A brokerage publishing “our agents are the best in every market” is the same self-serving signal one step removed, and AI is built to distrust exactly that. The signal that works is independent editorial validation, verified against real production and maintained over time. By definition, you cannot manufacture that about yourself.
What Invisibility Actually Costs the Number One Agent
Put real numbers on it. Ambuehl is the verified number one agent in Los Angeles: $315.85 million in sales across 33.5 transactions in 2024. That is an average sale price of roughly $9.4 million. At a conservative 2.5% commission, a single one of those transactions is worth about $235,000 in gross commission.
Now picture the buyer who just sold a company and is relocating to LA. He does what people now do. He asks ChatGPT for the best agent in Los Angeles, gets five names, and calls one off the Million Dollar Listing cast. That one AI answer just cost the most productive agent in the city a quarter of a million dollars, and she will never know the call existed.
It does not take many. If being invisible to AI costs Ambuehl a single deal a quarter, that is roughly $940,000 a year. One a month and it is north of $2.8 million. These are not leads slipping through a CRM. They are nine-figure-portfolio clients being routed, one query at a time, to agents who close a fraction of what she does, because AI cannot see that she outsells all of them.
And it compounds. Every month AI repeats the wrong names, it trains deeper on them. The longer the rightful number one stays invisible, the more permanent her replacement becomes.
This Is What EntitySeal Does
What AI rewards is a persistent, structured, independent record of who leads a market, verified against real production and maintained over time. That is the layer AI builds its answers from, and for most top agents and brokerages it sits empty. It is a different kind of asset than anything in a marketing department, and it is the asset EntitySeal builds.
Our optimization engine, EntityIQ, takes an agent’s verified production data and constructs a structured entity profile, then deploys it across an independent editorial authority network engineered for how AI systems read and cite entities. It is the reliable, maintained version of what worked by accident in those 11 markets, built deliberately for the agents and brokerages who should be winning the other 39. We monitor how AI cites your agents, catch the markets where it is naming the wrong people, and correct the record before the models train deeper on it.
That window matters. The brokerages that establish verified presence for their top producers now are the ones whose agents own the recommendation for years. The ones who wait will spend those years trying to unseat a television personality the model has already learned.
If your top producers are not the names AI gives, that is the problem we solve. Contact EntitySeal and we will show you exactly where your agents stand in AI search today and what it takes to fix it.
Frequently Asked Questions
Why does AI recommend real estate agents who aren’t the top producers?
AI assistants cannot access verified production data like RealTrends or the MLS. They build “best agent” answers from third-party information on the open web, weighting whatever is repeated most across independent sources. That favors agents with media coverage, high review counts, or aggressive self-published claims, not the agents who actually close the most homes.
How was this study conducted?
We compared AI’s “best real estate agent” recommendations across 50 top U.S. real estate markets against RealTrends verified rankings by sales volume and transaction sides. We measured how often AI surfaced the verified number one producer and how often it surfaced any of the verified top five.
Will updating my own website fix my AI visibility?
Largely no. An AirOps analysis found 85% of brand mentions in AI answers come from third-party domains and only 13.2% from a business’s own website. Self-published claims are the signal AI is built to discount. Visibility comes from verified, structured presence on the independent sources AI cites.
Can a brokerage fix this by issuing press releases?
No. A press release is a one-time, unstructured signal that AI mostly ignores and occasionally rewards by accident, and it decays quickly. AI also down-weights what a business says about itself, so a brokerage promoting its own agents is the signal AI distrusts most. The fix is independent, verified, structured presence maintained over time.
Why does AI get the answer right in some markets?
In the 11 markets where AI named the true top producer, a structured, data-driven third-party source existed that AI could read. That confirms the mechanism: AI is accurate when verified data about an agent exists in a form it can parse, and blind when it does not.