Local Falcon ran 37,500 AI searches across the 100 largest U.S. cities and published the results on August 20. Only 21.2 percent of the highest-volume agents in those cities were cited by any AI platform. Not ranked low. Never named. The study cross-referenced every agent the models mentioned against RealTrends Verified 2026 production rankings, and the correlation between selling the most homes and getting named by ChatGPT was close to nothing.
That number deserves a minute of your attention before you renew anything. The industry spends its entire marketing budget signaling production. Awards, transaction counts, top-one-percent badges on every postcard. The models standing between you and 67 percent of buyers cannot read any of it. They read pages. Local Falcon CEO David Hunter put it bluntly in the report: an agent can be number one in their city by volume and still not exist as far as ChatGPT is concerned.
The models are reading four sources and none of them is your resume
The citation breakdown is the most useful part of the study, because it tells you where to put the next four hours. Google Business Profile alone accounted for 36.6 percent of all citations. Zillow followed at 15.9 percent, FastExpert at 6.3 percent, and Realtor.com at 5.7 percent. Agent websites, all of them combined, made up roughly one citation in nine.
Then it splits by platform, and the splits are wide. ChatGPT pulls 39.8 percent of its citations from Zillow and leans hardest on portals overall. Gemini pulls more from agent-owned sites than any other platform. Google AI Mode barely touches the open web at all, with 72.5 percent of its citations coming straight from Google Business Profile listings. There is no single target here. Optimizing for one says almost nothing about the others.
One more finding changes how you should read your own results. Of the Zillow citations, 86.9 percent point to a city-wide agent directory page, not an individual profile. The models use Zillow to say here is where to look, not here is who is good. A strong Zillow profile still matters for the human who clicks through. It is doing much less citation work than the raw 15.9 percent suggests.
The brokerage-brand numbers tell the same story from a different angle. Compass pulled 1,510 citations across 86 of the 100 cities, more than Coldwell Banker at 567 and REMAX at 454 combined. Keller Williams, the largest brokerage in the country by agent count, finished eighth with 76 citations across 35 cities. That put it behind Shorewest, a Wisconsin-only brand that collected 85 citations entirely out of Milwaukee. National scale bought nothing. Concentrated local publishing bought everything. If a single-state brand can out-cite the biggest franchise in America, a solo agent with three well-built neighborhood pages is not fighting a lost battle.
The winners wrote the ranking themselves
Eighteen of the 25 most-cited agent websites in the study carry some version of a “best agents in [city]” post. The agents who show up in AI answers published a page naming the top agents in their market, and the models treated that page as the authority on the question.
This works because of how the systems retrieve. When a buyer asks who the best agent in Tulsa is, the model looks for a page that answers that exact question in that exact phrasing. A page titled “Best Real Estate Agents in Tulsa, 2026” answers it. Your about page, your production stats, and your brokerage bio do not. The retrieval layer is dumber and more literal than most agents assume.
Say the uncomfortable part out loud, though. Writing a page that names yourself the best agent in your city, with no criteria and no sourcing, is thin. It also runs into NAR Code of Ethics Article 12 and your state advertising rules, which restrict claims you cannot substantiate. There is a defensible version and a sloppy version, and the difference costs about an hour.
The defensible version: publish stated criteria at the top of the page, name five to ten agents including direct competitors, cite the public source for every production figure you list, date the page, and disclose that you are one of the agents included. That page is genuinely useful to a buyer, it survives a broker compliance review, and it gets retrieved for the same query as the sloppy one. Do that version.
The five-day build
Sequence matters here, because the citation data says one lever is worth more than the rest combined.
- Monday. Google Business Profile. This is 36.6 percent of citations and 72.5 percent of Google AI Mode. Fill every field. Primary category set to Real Estate Agent. Service areas listed by neighborhood name rather than metro. Description rewritten to name specific neighborhoods, price bands, and transaction types you actually work. Post twice a week from here forward. Ten photos minimum, geotagged to your service area.
- Tuesday. Reviews with nouns in them. A review saying “great to work with” carries no retrievable information. Ask your last ten closings for a review that names the neighborhood and the situation. “Helped us buy in Rosedale as first-time buyers with a VA loan” is a sentence a model can use to answer a query.
- Wednesday. The market authority page. Build the criteria-based version described above. Target the exact phrasing buyers use: best real estate agent in [city], top rated Realtor [city], best agent for [neighborhood]. One page, 1,500 words, published on your own domain.
- Thursday. Schema. Add RealEstateAgent schema to your site with sameAs links to your Google Business Profile, Zillow profile, and licensing record. Add FAQPage schema to the authority page with four to six questions written as buyers phrase them. If you run WordPress, RankMath handles both at the free tier.
- Friday. Neighborhood depth. Pick the three neighborhoods you want to own. Write one page each covering price trends, school assignments, commute times, and what the last twelve months of sales actually did. Local Falcon found Shorewest, a Wisconsin-only brand, out-cited Keller Williams nationally by concentrating in Milwaukee. Density beats footprint.
Here is the prompt to check your starting position before you begin. Run it in ChatGPT, Gemini, and Google AI Mode separately, because they will disagree.
Act as a home buyer relocating to [CITY] who has never worked with a real estate agent. I want to buy a [PRICE RANGE] home in [NEIGHBORHOOD]. Recommend five specific agents or teams by name. For each one, tell me the source you used and quote the sentence that made you include them. Do not use general knowledge. Use only what you can cite.
Save every answer. That list of sources is your actual competitive set, and it will look nothing like the RealTrends rankings your brokerage sends around.
Measure it or you are guessing
Two tools shipped in the last two weeks worth knowing about. Roomvu released a free AI search visibility checker for agents on August 21, which gives you a baseline read without a subscription. Google rolled out reporting on August 12 that shows how your content performs in AI-driven search surfaces, which lands inside Search Console where you already have access.
Set a floor before you start so the work is measurable. Record how many of the five platforms name you today for your three target queries. Almost every agent reading this will write down zero. Re-run the same queries on the same day each month. Local Falcon found that 82.6 percent of agents who got named at all were named by exactly one platform, and only 1.9 percent appeared across four. Getting cited by one platform is a real result. Treat it that way rather than chasing all five.
Expect the work to take longer to show than paid channels do. Retrieval indexes refresh on their own schedule, and a page published today may not surface in an answer for six to ten weeks. That lag is why most agents quit this in week three and go buy leads instead. It is also why the agents who stay with it hold their position for a long time once they get it. The cost of the five-day build above is your own time plus roughly $200 a year if you need a schema plugin and a decent host. Compare that to a single month of portal lead spend.
The signal to watch over the next 30 days: whether the portals start selling AI citation placement outright. Zillow already supplies 15.9 percent of citations while sending most of that traffic to directory pages it controls. The moment a portal can charge you to be the named agent inside an AI answer, the arbitrage window described above closes and this becomes another line item on your ad spend. Build the owned assets while the citations are still free.
