Real estate agents
The seller asks an assistant who knows their street. An answer can only say so of an agent if some page does.
A seller asks ChatGPT which agent sells the most condos in their neighborhood and how long it takes. A buyer asks for an agent who speaks Spanish and works with first-time buyers. The assistant answers from the pages it can read: the brokerage's, the team's, yours. They often disagree with each other.
The pages about an agent should state the same facts in the same words: the neighborhoods they work, the property types, the brokerage, the license, the languages, how long their listings take, what they charge where they choose to say it. Real estate had the highest no-schema rate of the forty trades we crawled, and an agent's facts are spread across more pages than most businesses' are.
How your customers ask
What sellers and buyers ask before they choose an agent
The question is about a place, a property and a timeline
Nobody asks an AI assistant for a great agent. They ask for the agent who sells condos in a named neighborhood, or who works with first-time buyers, or who can sell in sixty days because a job starts somewhere else. With web search on, the assistant reads the pages it can retrieve, and it can say those things about an agent only where some page states them.
Most of those questions turn on four facts: the neighborhoods you work, the property types you sell, what your listings take from list to contract, and your terms where you publish them. Languages and designations sit behind those. Your listing presentation answers all of it. Your website often answers none of it, and the brokerage's page about you answers a question nobody asked.
An agent also exists on more pages than most tradespeople do: the brokerage roster, the team site, a personal site, portal profiles, a social page. An answer can draw on several of them, and only facts that agree show a reader that they describe one person. When the brokerage page carries one phone number and the personal site another, or the team changed its name and one page did not, those pages describe two entities with one name, and an answer may repeat either one.
What every page about an agent has to state, identically
The list is short. The discipline is that it be the same list, in the same words, on every page that names you.
| The client asks | What your pages state | Form |
|---|---|---|
| Do you know my neighborhood | The neighborhoods and towns you work, named, with what you have sold there and when | Text on a page per area you can speak to; areaServed in the markup |
| Do you sell my kind of property | Condos, single-family, multifamily, land, new construction, luxury, investment: which, and your closed count for each | Text; your own numbers, counted from your own closings |
| How long will it take | What your listings took from list to contract over the last year, with the date you counted | A dated sentence, updated when you recount |
| What do you charge | Your commission or fee structure, where you choose to publish it, in words; otherwise how it is set | Text on the sellers page |
| Can we work in my language | Languages spoken by you and your team | Text on the about page; knowsLanguage on the Person node |
| Who are you, exactly | Legal name, brokerage, license number and state, one phone, one email, identical on your site, the team site and the brokerage roster | RealEstateAgent for the business; a Person node for you, tied to the brokerage page by sameAs |
| Are you any good | Client reviews quoted as text with the property type and the neighborhood | Quoted text, with the platform named |
RealEstateAgent is one of schema.org's LocalBusiness subtypes, so it carries address, telephone, opening hours and areaServed, the fields Google's local business documentation reads. You as an agent are a Person, and sameAs is the property that tells a machine the profile on the brokerage's site and the profile on yours are the same person. Google's structured data introduction recommends JSON-LD and asks that structured data describe only what a visitor can see.
Reviews go on the page as text, with the property type, the neighborhood and the platform they came from. Google's review snippet rules give an agency no review stars for reviews of itself on its own site and say not to aggregate ratings from other websites, so the quoted words do the work, not the review schema.
What the crawl found: real estate had the highest no-schema rate of forty trades
In our audit of 9,483 small business websites (40 trades, 90 metros across the US, Canada, Europe and the UAE, late August to early September 2026), 17.8% of all sites had no schema.org markup at all. Real estate agents had the highest rate of any trade, 33.6%, ahead of photographers at 32.7% and tax advisors at 31.0%; carpenters, the best of the trades with a hundred or more sites, were at 15.8%. Of the sites that did carry markup, most of it described the website rather than the business, and LocalBusiness schema of any subtype appeared on roughly a quarter.
Across the sample, 43.1% of sites contradicted themselves about who and where they are on their own three main pages, and 36.9% carried more than one phone number. An agent's cell, the office line, the brokerage's main number and a tracking number from a listing feed are four numbers, and a page that does not say which one to call leaves the machine to guess.
A crawler could read no review on 64.7% of the sites; 8.9% had rating schema. 61.5% had no question-and-answer content of any kind.
Our probe asked Claude, with web retrieval on, one buyer question for each of 9,240 businesses, and 14.9% of them came back named. Of every name it gave, 99.4% were actual businesses and 0.6% were directories or aggregators. One question, one assistant, one day: a direction, not a rate.
What to fix first, in order
- Reconcile the identity: legal name, brokerage, license number, phone and email, identical on your site, the team site and the brokerage roster. Ask the brokerage to correct its page; a roster entry usually has five fields, and those are the five.
- RealEstateAgent markup for the business and a Person node for you, with sameAs pointing at the brokerage page and your profiles, every value matching the page.
- A page per neighborhood you can speak to, with what you sold there and when, in text.
- Property types and the list-to-contract time for your own listings, counted and dated.
- Your terms, in words, where you publish them; otherwise a sentence on how they are set.
- Reviews on the page as text with the property type and the neighborhood, and no platform rating copied into the markup.
- Check what a crawler receives with the crawler access check, then measure the same buyer and seller questions before and after, as our methodology describes.
Go deeper
Articles for real estate agents and brokerages
Fundamentals
AI visibility for real estate agents: named, and under which brokerage
Brand & Positioning
Neighborhood and property-type pages built from an agent's own record
Authority & Mentions
What homeowners ask before choosing a listing agent, and what to publish
Measurement & Reporting
Measuring mentions for real estate agents: one afternoon, seller and buyer
What agents ask
The brokerage controls my page on its site. What can I do about it?
Ask for the fields to be corrected: a roster page usually carries a name, a phone, a license number and a photo, and those have to match yours. Then make your own site the page with the most facts, and point sameAs from your Person markup at the brokerage page, so a machine reading both knows they describe one agent rather than two.
Should I publish my commission?
That is your call and your brokerage's. What an assistant needs is an answer, not a number: if you publish your terms, publish them in words on your own site so they are quoted correctly; if you do not, say how they are set and that they are discussed at the listing appointment. A page that says nothing leaves the answer to another agent's page.
My leads come from a portal. Why does an assistant's answer matter?
Because the answer is mostly not a portal. Of every name Claude gave in our probe of 9,240 businesses, 99.4% were actual businesses and 0.6% were directories or aggregators. When a seller asks an assistant, they get agents by name, and which names depends on the pages about those agents. One assistant, one day: a direction, not a rate.
Do market statistics on my site help?
Only the ones you counted yourself, dated: what you sold, where, how long it took. Those are facts about you that exist on no other page, which is what makes them worth quoting. A market summary copied from a report exists on a thousand other pages and tells the assistant nothing about you.