September 6, 2026 · 7 min read
Under 10% of Agents Appear in AI Answers. How to Get Recommended by ChatGPT
Buyers ask ChatGPT for an agent and get six names. Fewer than 10% of agents ever appear. Here is the GEO plan that gets a brokerage into the answer, step by step.
A brokerage gets recommended by ChatGPT the same way it earns a spot in the Google 3-pack, only with stricter rules: one consistent name, address, and phone everywhere; a Google Business Profile with steady recent reviews; pages that answer specific local questions in the first paragraph; and a blog that publishes often enough that the models keep treating you as active.
FlyDragon's Q1 2026 buyer survey found 67% of buyers used an AI search tool (ChatGPT, Perplexity, Gemini, Claude, or Google AI Overviews) as a primary research step before contacting an agent, up from 17% about 18 months earlier. When those buyers ask for the best agent in a city, Omni Eclipse's 2026 report found ChatGPT names a median of six businesses. Six. And multiple 2026 studies find fewer than 10% of agents appear in AI answers to location-based questions at all.
That is a shortlist you are either on or not on, for every neighborhood you serve. The good news is that the models are not mysterious about how they choose. They are just strict.
How the models decide who gets named
An AI assistant answering "who is a good real estate agent in Decatur for a first-time buyer" is not searching the way Google searched in 2019. It is assembling an answer from sources it trusts, and it prefers sources that agree with each other. Four signals do most of the work:
Entity consistency. Your brokerage name, address, phone, and website must match exactly across your site, Google Business Profile, the MLS, portals, social bios, and directories. A suite number that appears on some listings and not others is a crack, and cracks get you left off the list of six.
Reviews, recent and specific. Models lean on review volume, recency, and the words inside the reviews. "Great agent" tells a model nothing. "Helped us relocate from Chicago to Brookhaven and knew the school zones" tells it exactly which question you answer.
Answer-first pages. Pages where the first paragraph contains a direct answer to a specific question, followed by structure the model can parse: headings, FAQ blocks, lists.
Freshness. A site that last published in March reads as dormant. A site that publishes ten to fifteen posts a month reads as a live business with current knowledge.
This is generative engine optimization, or GEO, and we covered the agent side in how real estate agents win in the age of AI answers. This post is the brokerage side: getting the whole entity recommended, not one lucky agent.
By the numbers: the AI front door
Veterans United's 2026 survey found 59% of buyers used at least one AI platform in their home search: ChatGPT 33%, Gemini 20%, Meta AI 10%.
Google AI Overviews now appear on an estimated 30 to 40% of queries (some trackers say roughly half) and on most local business searches, and they surface fewer businesses than the old map pack did.
Cotality's 2026 housing survey found 44% of buyers would pay more for a human to verify AI-generated information, and NAR's 2025 Profile of Home Buyers and Sellers found 88% of buyers purchased through an agent.
Get named in the answer, then be the human who verifies it.
The GEO checklist for a brokerage
Work through this in order. Most brokerages we audit fail items one and two.
Audit the entity. Pull your name, address, phone, and URL from your website, Google Business Profile, social bios, portals, and the MLS. Fix every mismatch. Do the same for each agent.
Fix the Google Business Profile. Whitespark's 2026 Local Search Ranking Factors put GBP signals at about 32% of local pack ranking, and AI Overviews draw on the same profile. Categories, service areas, hours, photos, weekly posts, and a Q&A section written in plain sentences.
Build one answer-first page per neighborhood you serve. Not a listings feed. A page that answers "what is it like to buy in X" in the first fifty words, then covers price ranges, commute, schools, and who on your team works there.
Add an FAQ block to every service page. Five to eight real questions, each answered in 40 to 80 words. Write them the way a buyer types them into ChatGPT.
Give every agent an entity page. Name, photo, neighborhoods, specialties, review excerpts, and a link to their GBP. Agents are entities too, and the brokerage page that lists them is how the model connects the names.
Run a review program. Ask at closing, ask with a link, ask for specifics. Track it monthly.
Publish 10 to 15 blog posts a month. Local, question-led, with an author name and a date on every one.
Interlink it all. Neighborhood pages link to agent pages, agent pages link to blog posts, blog posts link to the GBP.
What an answer-first page looks like
Say you serve Sandy Springs and want to be named when someone asks an assistant about relocating there with kids. The page is not titled "Sandy Springs Homes for Sale." It is titled "Moving to Sandy Springs With Kids: What Families Ask Before They Buy."
The first paragraph answers in about fifty words: which parts of the city families pick, the price range they see, and the commute reality. Then four short sections: schools and zones, the three most common neighborhoods, a typical closing timeline, and which two agents on the team have done the most relocation work there, with a sentence from one of their reviews. Then an FAQ block with six questions. At the bottom, links to the agent pages and two related blog posts.
That page gives a model everything it needs: a specific question, a direct answer, named people, third-party proof, and links that confirm the entity. Multiply it by every neighborhood and question type (downsizing, investing, first-time buying, luxury) and you have built what the models look for.
Reviews: the entity's credit score
Reviews are the signal brokerages most often neglect at the office level. Consumer research is consistent: 97% of consumers read reviews before choosing a local business, 47% will not consider a business with fewer than 20 reviews, and 73% only pay attention to reviews from the last month. Profiles with 100 or more reviews attract about 360% more website visits.
For an AI answer, recency matters as much as volume. A brokerage with 300 reviews and none since February looks less alive than one with 90 reviews and six from last month. So the program is monthly, not annual: every closing gets a request, every request asks for a specific detail, and the office reports review count and recency. We wrote about the downside of neglecting this in how one bad Google review can cost you a listing.
How Urban Marketing Edge handles this
GEO is built into how we run a brokerage's content, not sold as a separate service. In the Clarify step each month we choose a theme and the local questions that theme will answer, then the blog cadence (10 to 15 posts a month) turns each question into an answer-first post with an author, a date, an FAQ block, and internal links to neighborhood and agent pages. We audit and fix entity consistency in the first month, including every agent's profile, and the review request is built into the listing launch sequence so it happens at closing. Google Business Profile posts and Q&A are part of the weekly publishing schedule and go through the same post approval system as social. The monthly report shows GBP actions, website users, and review count and recency in plain language. Clareo Group's site reached 23.83K views and 14.56K visitors in 2025 on this approach. The 12 Week Program teaches the same website, SEO, and GEO workflow to an in-house marketer.
Questions brokers ask
How does ChatGPT choose which real estate agents to recommend?
It assembles an answer from sources it trusts and prefers sources that agree with each other. Consistent name, address, and phone across the web, recent and specific reviews, pages that answer local questions directly, and a site that publishes regularly all raise confidence. Omni Eclipse found it names a median of six businesses per city.
Can a brokerage really show up in AI search results?
Yes, and most do not, which is the opportunity. Fewer than 10% of agents appear in AI answers to location-based questions according to multiple 2026 studies. A brokerage with a clean entity, a review program, neighborhood pages, and a steady blog cadence competes against very few offices.
Do Google reviews affect AI recommendations?
Strongly. Reviews supply the volume, recency, and specific language that models use to decide whether a business is credible and relevant to a question. Since 73% of consumers only pay attention to reviews from the last month, aim for a monthly flow of specific reviews rather than one big push.
How often should a brokerage publish blog posts for AI visibility?
Ten to fifteen posts a month is the cadence we run for brokerages. Each post answers one specific local question in the first paragraph, carries an author and a date, and links to related neighborhood and agent pages. Freshness is a trust signal, and a dormant blog reads as a dormant business.
What is GEO in real estate marketing?
Generative engine optimization is the practice of structuring your web presence so AI assistants (ChatGPT, Gemini, Perplexity, Google AI Overviews) can find, trust, and cite you. It overlaps with local SEO but puts more weight on entity consistency, direct answers, FAQ blocks, and review recency.
Book a 30-minute strategy call
Six names per city. That is the whole contest, and right now most of your competitors are not entered. Getting in is a matter of entity hygiene, reviews, answer-first pages, and cadence: no big budget, but a real system.
Book a 30-minute strategy call and we will map the simplest plan for your content, themes, and monthly cadence, with AI visibility built into it. No pitch deck, no pressure. We will look at your entity consistency together, tell you honestly where the cracks are, and outline the first three pages to build. Plan options and pricing are on the plans page if you want to look ahead of time.
