AEO for real estate agents: why the top agents dominate AI recommendations

When a buyer asks ChatGPT or Gemini for a good agent, it names a few people and stops. If you are one of them, you have just been handed a lead that behaves like a referral, not a portal click. If you are not, the conversation happened without you. Most agents are invisible in AI search. Here is why the ones who show up win the lead, and how to become one of them.

Most agents are simply invisible in AI

Ask ChatGPT or Gemini to recommend a good buyer's agent in a specific town and watch what happens. You do not get a page of two hundred profiles the way you would on a portal. You get a short answer that names a few people, describes why each fits, and stops. For the agents named, that is a warm, pre-qualified introduction delivered at the exact moment someone is deciding who to trust with the biggest purchase of their life. For everyone else, the conversation happened and they were never in the room.

This is the uncomfortable truth of AEO for real estate. The habit has already moved. A 2025 Veterans United survey found that around 39% of prospective home buyers had used AI tools during their search, with ChatGPT the most common at roughly a third of respondents and Gemini next. Separately, BrightLocal's 2026 survey found 45% of consumers now use AI for local business recommendations, up from about 6% a year earlier, putting AI ahead of Yelp. Buyers are asking. The question is whether the model has anything confident to say about you.

And here is the part most agents miss: being invisible in AI is not a small marketing gap. It is a gap at the most valuable point in the funnel, because the lead that arrives through an AI recommendation behaves nothing like the lead that arrives through a portal.

Curious how AI engines describe your brand right now? Get a free visibility audit and see where you stand across ChatGPT, Gemini and Perplexity.

Why the AI-sourced lead is worth more

The economics of real estate leads are brutal and well documented. NAR data, widely cited across the industry, puts the conversion rate on cold online portal leads at roughly 0.4% to 1.2%. Top agents with disciplined follow-up push that to maybe 3% to 5%. In plain terms, an agent buying portal leads may need somewhere between eighty and two hundred and fifty of them to close a single deal. That is the model Zillow and Google Ads leads run on: volume, speed-to-lead, and a lot of leakage.

A referral converts on a completely different scale, because the trust arrives before the conversation does. When a friend says "use this agent", the buyer is not comparison-shopping twelve strangers; they are looking for a reason to say yes. Sources across the industry report referral-style conversion many times higher than cold online leads. An AI recommendation is structurally closer to a referral than to a portal lead. The buyer asked an assistant they trust, and the assistant named you with a reason attached. You are not one of two hundred; you are one of three, endorsed.

Some vendor blogs put eye-catching numbers on this - specific close rates, specific commission-per-lead figures. We are deliberately not repeating those, because they trace back to single marketing sources with no published methodology. The verified, defensible claim is enough on its own: cold online leads convert in the low single digits at best, warm recommendation-style leads convert far higher, and an AI recommendation sits firmly in the second category. That is why showing up is worth so much more than showing up somewhere on a portal.

"On a portal you are one of two hundred. In an AI answer you are one of three, with a reason attached."

How LLMs actually decide which agent to name

To win the slot you have to understand where the model is reading. When an assistant answers "best buyer's agent in [town]", it is not running a live MLS query. It is drawing on what it has learned and, in browsing modes, what it can pull in that moment: your own website, your profiles on Zillow and Realtor.com, your Google Business Profile, review platforms, local "best agents" roundups, press mentions, and your social footprint. It assembles a picture from all of those and then decides who it can describe with confidence.

That word, confidence, is the whole game. A model will not stake a recommendation on a fuzzy, contradictory identity. It reaches for the agent whose story is consistent everywhere it looks: the same name, the same brokerage, the same service area, the same specialism, corroborated by recent reviews and a clear website. The agent who is well known offline but barely described online loses to the agent whose signals are clean and aligned, even if the second agent sells fewer houses. The model recommends what it can read clearly, not necessarily who is best.

Two agents in the same town can therefore get wildly different treatment. One has a Zillow profile with forty recent reviews, a website that says plainly "buyer's agent specialising in first-time buyers in [neighbourhood]", and matching details on Google. The other has a stale profile, a website that says "your trusted local expert" and nothing specific, and reviews that stopped two years ago. The model names the first and never mentions the second. Neither knows why.

The identity signals that actually move the needle

Getting named is not mysterious once you see it as an identity problem rather than a ranking problem. A handful of signals do most of the work, and almost all of them are within your control.

• A consistent professional identity everywhere. Your name, brokerage, licence area, phone, and specialism must be identical across your site, Zillow, Realtor.com, Google, and LinkedIn. Every contradiction - a different service area here, an old brokerage there - lowers the model's confidence at the exact moment it is choosing who to name.
• Recent, specific reviews. Volume matters less than freshness and detail. Twenty reviews from this year that mention "first-time buyer", "negotiated below asking", "closed in three weeks" tell a model far more than two hundred generic five-star ratings that stop eighteen months ago. Specific praise maps directly onto the specific questions buyers ask.
• A stated specialism and service area. "Your trusted local expert" is invisible to a model. "Buyer's agent for first-time buyers in [neighbourhood], [city]" can be matched to a real query. Say what you do and for whom, in plain words, on your own site.
• Presence in the sources models trust for agents. Zillow and Realtor.com profiles, Google Business Profile, and credible local "best agents in [city]" roundups are exactly where assistants look. Being genuinely and consistently present across them is corroboration; being absent is a blank the model fills with someone else.

None of this is a growth hack. It is the same trust an experienced agent builds offline, made legible to a machine that reads before it recommends.

Structured data and local markup: making yourself machine-readable

There is a technical layer under all of this that most agents ignore, and it is quietly decisive. Models and the crawlers that feed them read structured data far more reliably than they read marketing prose. Schema markup on your own site turns "your trusted local expert" into unambiguous, labelled facts a machine can lift without guessing.

For an individual agent, the practical pieces are a RealEstateAgent or Person schema stating your name, brokerage, and role; LocalBusiness markup with your service area, address, and hours; and AggregateRating or Review markup that exposes your genuine reviews in a form a model can parse. Done properly, this means that when an assistant assembles its picture of you, it is working from clean structured facts rather than trying to infer them from a hero image and a tagline. You are removing ambiguity, and ambiguity is the thing that gets you skipped.

The same discipline applies to your listings and area pages. A page about "homes for sale in [neighbourhood]" with proper markup, a clear author, and a plainly stated local specialism does double duty: it helps buyers and it feeds the model a well-labelled example of your expertise in that exact area. The goal throughout is not to trick anything. It is to make the true story of your business impossible to misread.

Why the top agents pull away

Put the pieces together and you can see why AI recommendations compound advantage for the agents already doing well. Success offline generates the very signals models reward: more closings mean more recent reviews, a clearer track record, more local mentions, a more established profile. Those signals make the model more confident, so it names you more often, which brings more warm leads, which produce more closings and more reviews. The agent who is already winning becomes the agent AI keeps recommending.

The good news for everyone else is that the loop is enterable, because most of what feeds it is signal quality, not signal volume. An agent with a modest but clean, recent, consistent, well-marked-up presence can outrank a busier agent whose online identity is a contradictory mess. You do not need to be the biggest name in town. You need to be the clearest.

That is the real reframing. The old game was ranking on a page with room for everyone. The new game hands out three names, and it hands them to the agents it can describe with confidence. The buyers have already changed how they ask. The only question left is whether, when they ask, the answer includes you.

Where to start this week

Do not try to fix everything at once. Start by finding out what the assistants actually say when a buyer asks for an agent in your market. Type your real prospects' questions into ChatGPT, Gemini, and Perplexity - "best buyer's agent in [your town]", "who should I use to sell my house in [your area]" - and note two things: whether you appear, and who appears instead.

Whoever keeps getting named is your lesson. Look at what they have that you do not: cleaner profiles, fresher reviews, a clearer specialism, a website that states plainly what they do. That gap, ranked by the agents currently taking your slot, is your to-do list. Then fix the identity basics first - consistency across every profile, a steady flow of recent reviews, a plainly stated specialism and service area - because those move the needle fastest. The markup and the deeper content follow. The agents who win the AI lead are rarely the best at selling houses. They are the ones the model can read.