GEO for Real Estate: How Realtors and Loan Officers Get Named Inside AI Answers
Short answer: Generative Engine Optimization, or GEO, is the work of making AI search tools identify you correctly, place you in the right market, and name you inside the answer they give. SEO earns a ranking. GEO earns a recommendation. In 2026 the recommendation happens before anyone clicks anything, and the list is short: ChatGPT names a median of six agents per city when someone asks who to hire.
Source: RECSC 2026 Raleigh AI Visibility Benchmark, 75 Raleigh-area real estate and mortgage brands tested in clean, non-personalized AI sessions.
What changed since this playbook first went up
A lot moved in one month. Here is what changed, and where each change shows up below.
- Reddit stopped being a shortcut. Reddit citations inside ChatGPT search dropped 86 percent in about four days around August 20. See the two-track play.
- ChatGPT runs its own index now. Bing shows up mainly in deep research, so "optimize Bing" is no longer a ChatGPT plan. See how the machine decides.
- Ranking top ten buys less in AI Overviews. The share of AI Overview citations coming from top-ten pages fell from 76 percent to 38 percent. Topic coverage now does more of the work.
- For "who should I hire," the sources are yours. 98.1 percent of the sources ChatGPT cited on real estate agent questions were the business's own website or Google Business Profile. Zillow got zero.
- FAQ rich results are gone. Google removed them on May 7, 2026. The markup still gets read. See the code playbook.
- You can finally measure some of this in Google. Search Console now has a generative AI performance report. See the scoreboard.
- New section: exactly how Realtors and loan officers should build answer pages that get cited. Jump to answer pages.
Here is the part nobody says out loud at the office meeting.
A buyer relocating to Raleigh from New Jersey does not open Google, scroll ten blue links, and compare agent websites like it is 2015. They open ChatGPT and type something like "we are moving to the Triangle in the spring with two kids, good schools, under 700k, who should we talk to." Then they get a paragraph. Sometimes with names in it. Usually a short list. A 2026 study of 334 ChatGPT answers across 197 markets put the median at six agents per city.
That paragraph is the new first showing. You are either in it or you are not, and you never find out which.
Same thing is happening on the mortgage side, and faster. Realtors vet loan officers with AI now. Borrowers ask AI whether a rate they were quoted is reasonable and who else they should call. If the tool cannot confidently place you in the market you work in, you do not make the list. Not because you are bad. Because you are unresolvable.
The shift in one sentence
For twenty years the goal was to rank. The goal now is to be the source the answer is built from, because most people will never scroll past the answer.
And Google is pushing harder in that direction every month. It is steering more searches into AI Mode, where one analysis of Similarweb data found 93 percent of searches end without a click.
This is not a reason to panic and it is definitely not a reason to buy an AI tool subscription. It is a reason to fix your signals. That is the whole playbook below.
Definitions that actually matter
What is the difference between SEO, AEO, and GEO?
People use these acronyms interchangeably and then wonder why their strategy is a mess. They are not the same work, they do not use the same metrics, and you need all three running at once.
SEO: Search Engine Optimization
Getting your pages to rank in a list of links so someone clicks through to your site. Still the foundation. Still how Google decides who is credible enough to pull into an AI answer. Not dead. Just no longer the finish line.
AEO: Answer Engine Optimization
Getting your content pulled into featured snippets, voice results, and "people also ask" boxes. This is where the answer-first writing habit came from, and it is the bridge skill between old SEO and everything AI is doing now.
GEO: Generative Engine Optimization
Getting your name, your market, and your expertise cited accurately inside an AI-generated answer. Content plus code plus proof plus third-party corroboration. The 2026 job, and the one almost no agent or lender is doing on purpose.
The plain-English version
SEO gets you found. AEO gets you quoted. GEO gets you recommended.
The RECSC framework
Every AI answer puts you through three gates. Most agents fail at gate one.
When we ran the 2026 Raleigh AI Visibility Benchmark across 75 real estate and mortgage brands, the failures were not random. They stacked in a specific order. You cannot get recommended if AI cannot place you. You cannot get placed if AI cannot identify you. That is the whole hierarchy.
Gate 01 · Resolution: does AI know who you are?
Ask an AI tool about you by exact name. Does it describe you, describe a stranger with your name, or blend three people into one confident and completely wrong bio?
This is entity resolution, and it is the gate almost nobody checks. If your name does not resolve to one clear professional entity, nothing else you do matters.
16 of 75 brands failed this outright.
Gate 02 · Placement: does AI know where you work?
You can be correctly identified and still get filed under the wrong city. Multi-state licensing, a corporate umbrella site, an old brokerage bio, a team page from four years ago. AI cross-references all of it and makes a call.
Loan officers get hit hardest here because the license trail is public and the personal web presence usually is not.
37 of 75 had weak market signals. 12 of 25 loan officers were placed outside their market.
Gate 03 · Recommendation: does AI name you unprompted?
The real test. Not "who is Emily Wyatt," but "who should I hire in Raleigh." No brand name in the prompt. Pure discovery, which is how actual buyers and borrowers ask.
This gate rewards proof: reviews with substance, local press, community presence, third-party mentions, content other people cite, and, for local hiring questions, a Google Business Profile and website that say exactly the same thing about you.
Only 22 of 75 appeared in that evidence at all.
Read that again
Seventy percent of the brands we tested were invisible in open discovery. Not outranked. Invisible. That is not a competitive market, that is an empty room with the door propped open.
How the machine actually decides Updated
How does AI decide which real estate agent or loan officer to name?
AI does not remember you. It looks you up, then argues with itself. The technical name is Retrieval-Augmented Generation. Skip the acronym, keep the concept, because it explains every strange thing you have seen an AI tool say about your business.
A language model on its own is a closed book with a knowledge cutoff. It cannot know about your new listing, your brokerage change, or your market. So modern AI search does two separate things in sequence.
Step one, retrieval. The tool goes and finds source material about the question. Your website, your Google Business Profile, local data providers like Foursquare, Apple Business Connect, review sites, directories, Reddit threads, local news, LinkedIn, your old brokerage page that you forgot exists. This step is mechanical. It is about whether your content is structured clearly enough to be found and pulled in chunks.
Step two, generation. The tool reads those sources, decides which ones to trust, and writes an answer. This step is a judgment call. It is about credibility, consistency, and whether your sources agree with each other.
What changed in step one since August
- ChatGPT built its own search index. It now pulls from several providers, with Bing showing up mainly in deep research. Bing visibility is no longer a stand-in for ChatGPT visibility.
- Local data providers matter more than most agents know. One August report found Foursquare data feeding 60 to 70 percent of ChatGPT's local search results. Almost no agent or loan officer has ever logged in to theirs.
- Most AI crawlers do not run JavaScript. They read the raw HTML and leave. IDX widgets, testimonial sliders, text hidden in tabs, and words baked into graphics often never make it into step one at all.
Why this matters for your budget
These are two different battles that need two different investments. Clean structure with no credibility gets ingested and ignored. Strong credibility with messy structure never gets retrieved in the first place. Most Realtor websites fail both. Most loan officer web presences fail the first one before the conversation even starts.
It also explains the thing that frustrates agents most: why AI sometimes says something outdated or flatly wrong about you. It is not making things up out of nowhere. It is resolving a conflict between sources that disagree, and it resolves that conflict without calling you first.
Which means the fix is not "post more." The fix is making your sources stop contradicting each other. That is exactly what a Google + AI Visibility Audit is built to find.
The two-track play Updated
Why does what wins in Google AI Overviews lose in ChatGPT?
This is still the most useful strategic frame in the playbook, and it still gets ignored constantly. "AI search" is not one thing. Google's AI answers and standalone tools like ChatGPT and Perplexity pull from different places. One strategy cannot serve both. What changed this summer is where each track pulls from.
Track 1 · Rank first, then cover the topic: winning Google AI Overviews
When this playbook first went up, the go-to stat was that roughly 81 percent of AI Overviews pulled at least one top-ten organic result, with about a one in three chance for the number one result specifically. That relationship has loosened. Ahrefs analyzed 863,000 keywords and found the share of AI Overview citations coming from top-ten pages fell from 76 percent in July 2025 to 38 percent by March 2026. The rest came from positions 11 to 100 and beyond.
The reason is query fan-out. Google splits one question into several related sub-questions, then cites pages that answer those, even when they do not rank for the original query. Ahrefs notes that better citation detection explains part of the drop, so hold the exact number loosely. The direction is not in question.
Translation for you: classic SEO is still the entry fee. If you are not ranking, you are not getting cited. But one strong page is no longer enough. You need the page, plus the related questions around it, answered clearly on pages that link to each other.
- Real local pages with real local substance, not one page with twelve city names swapped in.
- A Google Business Profile that is fully built, categorized correctly, and posted to consistently.
- Neighborhood, school, relocation, and price-band content that answers the actual question in the actual language people use.
- Internal linking that connects your service pages to your market pages so Google understands the relationship.
- New: answer pages for the follow-up questions, built with the structure in the answer pages section below.
- New: short video answers. YouTube makes up 5.6 percent of AI Overview citations and grew 34 percent in six months, per the same Ahrefs research.
- New: a spot on the credible local lists. In August, Google's AI Overviews were caught quoting third-party "best of" listicles for local businesses. You want to be named on the good ones. Never pay to be placed on a junk one.
Track 2 · Owned first, corroborated second: winning ChatGPT and Perplexity
Different animal entirely. When this playbook first went up, the standard advice was that standalone AI tools pulled most of their citations from outside Google's top results, leaning on encyclopedic references and community sources: Reddit, Quora, forums, review platforms, and content written by humans who are not selling anything. That is still partly true for broad informational questions. Two things changed.
First, the Reddit shortcut broke. Around August 20, 2026, Reddit citations inside ChatGPT search dropped 86 percent in roughly four days, according to Promptwatch data reported by HousingWire. Google, for its part, says Reddit gets no special preference in its rankings.
Second, local hiring questions run on your own assets. A 2026 study of 334 ChatGPT answers across 197 markets found 98.1 percent of cited sources were the business's own website or Google Business Profile. Google Business Profile and Maps alone were 57.6 percent. Zillow, Realtor.com, Redfin, and Trulia got zero citations across all 1,567 links.
Translation for you: for "who should I hire," your owned assets carry the answer. Everything else is what makes AI believe them.
- A Google Business Profile and a website that say the exact same thing about who you are, where you work, and what you specialize in.
- Reviews that name your city, your neighborhood, and the specific transaction type.
- Claimed, matching listings on the local data sources AI reads: Foursquare, Apple Business Connect, Yelp, and Facebook.
- Local press, podcast appearances, association bios, chamber listings, and builder or lender partner pages.
- Genuine participation in local community threads and relocation forums, helpful and not pitchy, as supporting evidence rather than the whole plan.
- Consistent name, title, market, and specialty everywhere your name appears, so nothing contradicts anything.
The uncomfortable part
Track 2 is the one real estate professionals hate, because it cannot be outsourced to a scheduling tool and it does not produce a pretty dashboard. It is also where the least competition is. Make of that what you will.
The August Reddit drop is the lesson. Visibility you rent can vanish when a platform changes one setting. Visibility you own, backed by real corroboration, does not.
Playbook one: content
How should real estate content be written so AI can quote it?
Write so a machine can lift one paragraph out and have it still make sense. AI does not read your blog post like a person does. It breaks content into chunks and pulls the chunk that answers the question. If your best point only makes sense after four paragraphs of throat-clearing, it will never get pulled.
Answer first, always
Every page and every post gets a direct 40 to 60 word answer near the top, before the story, before the setup. If the post is "how much are closing costs in Wake County," the first block of text says what they are. Then you earn the rest of the read.
Build in modules, not narratives
Each H2 section should be able to survive being copied out of context. Question as the heading, answer immediately underneath, self-contained. That is what makes a page citable in eleven different answers instead of one.
Answer the next question too New
Because Google now fans one question out into several, the page that wins is the one that already answered the follow-ups. After "how much are closing costs," a buyer asks who pays them, when they are due, and whether the seller can cover them. Give each of those its own H2. That is topic coverage, and it is what Track 1 now rewards.
Short sentences. Real words.
Aim for 15 to 20 words per sentence and 2 to 4 sentences per paragraph. Cut hedging language like "might," "could potentially," and "some experts suggest." AI is filtering for confident, specific, verifiable claims. Vague copy reads as low value to a machine and to a buyer.
Use the words your clients use
Nobody types "residential acquisition opportunities in the Triangle region." They type "is Wake Forest a good place to buy right now" and "what credit score do I need for a mortgage in NC." Write the page around the question, not around the keyword you wish existed.
Skip complicated tables
Parsers read linearly, top to bottom. A table's meaning lives in the relationship between rows and columns, and that relationship gets mangled on the way in. If you have comparison data, format it as a labeled list with clear key and value pairs. Cleaner for AI, better on mobile, which is where your buyers are anyway.
Put the words in the page, not in a widget New
If your best content lives in an IDX feed, a review widget, a rotating slider, a PDF market report, or a Canva graphic, most AI crawlers never see it. Write your neighborhood notes, your reviews, and your market read as real text on the page. If a bot cannot copy and paste it, it cannot quote you.
Publish the thing only you could publish
Your original market read. Your actual numbers. What you saw at eleven showings last month. That is the content AI cannot generate from thin air, and it is the content that makes you a source instead of a summary. Build it into a repeatable system with the Market Authority Engine rather than doing it whenever inspiration strikes, because inspiration will not strike in March when you are busy.
Playbook two: the code Updated
Which schema and code changes help a real estate website get cited by AI?
Schema is how you stop asking AI to guess. Your content says what you do. Your markup proves it in a format machines cannot misread. This is the least glamorous work in the entire playbook and it is the fastest thing to fix on most Realtor websites.
Clean semantic HTML first
One H1 per page and one only. On Squarespace that is your page or post title field, which is exactly why you should never add a second H1 inside a code block. Main content in the body of the page, navigation in nav, sidebars and related-post modules in aside so AI knows what to skip. Headings in actual order. It sounds basic. Half the agent sites we audit fail it.
Make sure the words are actually in the HTML New
Here is the 30-second test. Open your site on a computer, right-click, choose View Page Source, then press Ctrl+F and search for a sentence from your homepage. If it is not in the source, most AI crawlers cannot read it either. Fix that before you touch anything else in this section.
Do not block the bots you want to be found by New
Check your robots.txt file. Search crawlers like OAI-SearchBot (ChatGPT search), PerplexityBot, Claude-SearchBot, and Googlebot should be allowed. Blocking GPTBot, which collects training data, is a separate decision from blocking ChatGPT search. Plenty of templates and well-meaning developers block both without asking.
The schema types that carry weight for real estate and mortgage
- RealEstateAgent or LocalBusiness. Your name, market, service area, address, phone, and hours, stated in machine-readable form instead of implied by a footer graphic.
- Person. You as an entity, with job title, employer, credentials, and social profiles linked. This is the single most underused markup in the industry and it feeds directly into gate one. Loan officers: add your NMLS ID as a credential.
- FinancialService. For mortgage brokers and branches, the business entity alongside your Person markup, with the states you are licensed in and the area you serve.
- Service. Each thing you actually sell, described distinctly. Listing marketing is not the same as buyer representation, and relocation support is not the same as either.
- FAQPage. Atomic question and answer pairs, pre-packaged in exactly the format retrieval systems want.
- Review and AggregateRating. Where legitimately applicable. Social proof a machine can count.
- Article and BreadcrumbList. For blog content, so your posts are understood as authored work inside a structured site instead of floating text.
Myth worth killing, updated
"Google got rid of FAQ schema, so stop using it." Here is what actually happened. On May 7, 2026, Google stopped showing FAQ rich results in search, and it later dropped FAQ support from the Rich Results Test. It did not stop reading the markup. Google still parses FAQ structured data to understand a page, and FAQPage markup still hands AI retrieval systems clean question and answer pairs. Keep it. Just stop expecting the accordion to show up in the results.
New myth worth killing
"Add an llms.txt file and AI will find you." Ahrefs looked at 137,210 domains in May 2026. 97 percent of llms.txt files were never fetched by anything, and AI bots never went looking for the ones that did not exist. It is harmless. It is not a strategy.
If your site was built on a template and you have never touched the code, this is where a Realtor Authority Website System build pays for itself. A website that cannot be parsed is a brochure with a domain name attached.
GEO for Realtors and mortgage pros New
How should Realtors and loan officers structure answer pages to get cited by AI?
Build each page around one real client question. Put the question in the H1, answer it in 40 to 60 words directly underneath, then cover the follow-up questions in their own H2 sections. Add local proof only you have, finish with five to eight FAQs, link the page to your service and market pages, and mark it up with Article, FAQPage, Person, and BreadcrumbList schema.
An answer page is a page built to be lifted. Every section can be quoted on its own, and the whole page tells AI exactly who is answering, where they work, and why they would know. Most agent and lender websites have zero of these. They have a home page, an about page, and a contact form. That is a business card, not a source.
The anatomy of an answer page
- The question as the H1. Written the way a client says it, with the place in it. "How much are closing costs for buyers in Wake County?" beats "Closing Cost Guide" every time. On Squarespace, the page or post title is your H1. Never add a second one.
- The answer block, directly under the H1. 40 to 60 words. Lead with the number, the range, or the yes or no, then the one condition that changes it. This is the paragraph AI lifts, so it has to make sense with nothing around it.
- A byline and a visible updated date. Your name, title, license or NMLS number, and market, plus "Updated September 2026." Authorship and freshness are both trust signals, and the date needs to be on the page, not only in the code.
- A key facts summary. Three to five short labeled lines, like "Typical range," "Who usually pays," and "When it is due." Labeled lines, not a table.
- H2s written as the follow-up questions. This is the fan-out coverage. Each H2 is a real question, and the first sentence under it answers that question outright.
- A local proof section. What you have actually seen. Your numbers, your transactions, the neighborhood detail a national site could never write. This is your E-E-A-T, on the page.
- An FAQ section. Five to eight questions pulled from real calls, texts, and emails. Each answer is two to four sentences and stands alone.
- Internal links in the body. Up to your service page, sideways to two or three related answer pages, and over to your market page. Descriptive anchor text, never "click here."
- One clear next step. The call to action should match the question. A closing cost page offers a buyer net sheet. A credit score page offers a pre-approval conversation.
- Schema in the page header. Article or BlogPosting, FAQPage, Person, BreadcrumbList, and your RealEstateAgent or FinancialService entity. On Squarespace, this goes in the post's page header code injection.
What an answer page looks like for a Realtor
This is an example of the structure. Swap in your own verified local numbers before you publish anything like it.
- H1How much are closing costs for buyers in Wake County, NC?
- Answer blockBuyers in Wake County typically pay 2 to 5 percent of the purchase price in closing costs, depending on the loan type, lender fees, and prepaid taxes and insurance. On a $450,000 home, that is roughly $9,000 to $22,500. Seller concessions can cover some or all of it if you negotiate for them in the offer.
- Follow-up H2sWhat is included in buyer closing costs in NC? · Who pays closing costs in North Carolina, the buyer or the seller? · How do seller concessions work right now in Wake County? · When do I pay closing costs, and how?
- Local proof H2What I have seen on Wake County closings this year
- Internal linksUp to your buyer representation page. Sideways to "How much house can I afford in Raleigh?" and "What is a due diligence fee in NC?" Over to your Wake Forest or Apex market page.
- Next stepRequest a personalized buyer net sheet.
What an answer page looks like for a loan officer
Same structure, with mortgage compliance built in. Again, verify every program detail with your own guidelines before publishing.
- H1What credit score do I need to buy a house in North Carolina?
- Answer blockMost conventional loans need a 620 minimum. FHA loans can go as low as 580 with 3.5 percent down. VA and USDA have no official minimum, but most lenders look for 580 to 640. A higher score lowers your rate and your mortgage insurance, so even 20 points can change your payment.
- Follow-up H2sCan I get a mortgage in NC with a 580 credit score? · How much does my credit score change my rate? · How fast can I raise my score before I apply? · Does getting pre-approved hurt my credit?
- Local proof H2What I am seeing with Triangle first-time buyers this year
- Internal linksUp to your pre-approval page. Sideways to "FHA vs. conventional in NC" and "How much house can I afford?" Over to your Raleigh or Triangle market page, and to your Realtor partner page.
- Next stepStart a pre-approval conversation.
For loan officers: compliance that doubles as a GEO signal
- Your NMLS ID and your company's NMLS ID on the page and in your Person schema. It is required, and it is also one of the strongest entity signals you have.
- The states you are licensed in, stated plainly, plus the metro you actually serve. This is the Gate 2 fix. Remember, 12 of 25 loan officer brands in our benchmark were placed outside their market.
- An Equal Housing Opportunity statement.
- Dated rate and program language. "As of September 2026," with no guarantees.
- Nothing that reads like a referral-for-value arrangement with agents or builders. Co-marketing has RESPA rules. Get compliance sign-off before you publish partner content.
The internal linking pattern: hub and spoke
- The hub is your service page: Buyer Representation, Listing Services, or Pre-Approval.
- The spokes are 6 to 12 answer pages, one question each.
- Every spoke links up to the hub in the first third of the page.
- Every spoke links sideways to two or three sibling questions.
- The hub links down to every spoke, using the question as the anchor text.
- Market and neighborhood pages link to the spokes that matter most in that market.
- Breadcrumbs on every spoke, like Home › Buyers › Closing Costs in Wake County.
This is what turns a dozen answer pages into topic coverage, which is exactly what AI Overviews now reward. One page answers a question. A linked cluster makes you the source on the subject.
FAQ rules that get sections lifted
- Use real questions in real wording, pulled from your inbox and your phone, not from a keyword tool.
- Answer in the first sentence. Explain after.
- Keep each answer to two to four sentences. If the honest answer is "it depends," say what it depends on.
- Do not repeat your H2s. The FAQ catches the leftover questions.
- Make the FAQPage markup match the visible text word for word.
The schema stack for an answer page
Here is a trimmed example of the markup that goes in the page header. Replace everything in brackets, and make sure every FAQ in the markup appears on the page exactly as written.
{
"@context": "https://schema.org",
"@graph": [
{
"@type": "BlogPosting",
"headline": "How much are closing costs for buyers in Wake County, NC?",
"datePublished": "[2026-09-01]",
"dateModified": "[2026-09-21]",
"author": { "@id": "[https://yoursite.com/about#person]" }
},
{
"@type": "Person",
"@id": "[https://yoursite.com/about#person]",
"name": "[Your Name]",
"jobTitle": "[Realtor or Loan Officer]",
"worksFor": { "@type": "[RealEstateAgent or FinancialService]", "name": "[Business]" },
"areaServed": "[Raleigh, NC]",
"hasCredential": { "@type": "EducationalOccupationalCredential", "credentialCategory": "license", "name": "[License or NMLS #]" },
"sameAs": ["[Google Business Profile URL]", "[LinkedIn URL]"]
},
{
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "[Question exactly as it appears on the page]",
"acceptedAnswer": { "@type": "Answer", "text": "[Answer exactly as it appears]" }
}
]
},
{
"@type": "BreadcrumbList",
"itemListElement": [
{ "@type": "ListItem", "position": 1, "name": "Home", "item": "[https://yoursite.com/]" },
{ "@type": "ListItem", "position": 2, "name": "Buyers", "item": "[https://yoursite.com/buyers]" }
]
}
]
}Answer page checklist
- One real client question, with the place in it, as the H1.
- A 40 to 60 word answer directly underneath.
- Byline with license or NMLS number and a visible updated date.
- Three to five labeled key facts.
- Follow-up questions as H2s, each answered in its first sentence.
- A local proof section only you could write.
- Five to eight FAQs with matching FAQPage markup.
- Links up to the hub, sideways to siblings, over to the market page.
- One next step that matches the question.
- All of it readable in View Page Source.
Playbook three: proof Updated
Why does first-hand experience matter so much for AI visibility?
Experience is the one thing AI cannot fake, and you already have it. AI's biggest liability is confidently saying something false. So these systems filter hard for sources that look like ground truth. Google calls the framework E-E-A-T: Experience, Expertise, Authoritativeness, Trustworthiness. For real estate and mortgage professionals, the first E is the unfair advantage nobody is using.
You have sat at closing tables. You have seen an appraisal come in 40k light and watched a deal get saved. You know which Wake Forest neighborhood floods and which builder's punch list actually gets finished. A language model has none of that. It can only repeat what someone published.
So publish it.
- Write in first person with specifics. Not "buyers often face inspection issues." Try "I have had three deals in Rolesville this year where the inspection turned up the same crawl space problem, and here is what we did." Specificity is the credibility signal.
- Publish original data. Your own numbers from your own market. Days on market for the price band you actually work. What percentage of your buyers came from out of state. Original data gets cited, and citations compound.
- Use real photos and video. Your listings, your walkthroughs, your face. Stock photography of a generic kitchen signals nothing to anyone.
- Build a real author bio. Credentials, market, years, specialties, license number, linked profiles, marked up with Person schema. This is what resolves you as an entity.
- Coach your reviews for specifics, never for ratings. "Great to work with, five stars" tells a machine nothing. A review that says "helped us relocate from Ohio to Apex and found us a builder before we ever flew in" is a placement signal, a specialty signal, and a trust signal in one sentence. Ask every client, not just the happy ones, and never offer anything in exchange. The FTC's rule on fake and incentivized reviews carries penalties of up to $53,088 per violation.
- Keep reviews coming, and claim every profile. New One analysis of AI traffic to agent review profiles, from May to June 2026, found reviews from the last 30 days drew 40 percent more AI reads than reviews five or more years old. Profiles with 100 or more reviews were read 2.3 times more than those under 50. Unclaimed profiles were read 27 percent less than claimed ones with the same review count.
The small operator advantage is real here. A national brokerage cannot publish a first-person crawl space story from Rolesville. You can do it this afternoon.
Scoreboard Updated
How do you measure AI visibility when the clicks are not coming?
Traffic reports are getting worse and it is not always your fault. When AI answers the question in the results, the click never happens. Early 2026 data from SparkToro and Similarweb put zero-click Google searches at 68 percent overall, 83 percent when an AI Overview appears, and 93 percent in AI Mode. Your visibility can grow while your sessions drop. If you only measure sessions, you will kill the thing that is working.
Track these instead:
- Inclusion rate. Pick ten questions your ideal client actually asks. Run them monthly in clean sessions across ChatGPT, Gemini, Perplexity, and Google AI results. Count how often you are named. That is your new ranking report.
- Entity accuracy. Ask AI who you are and where you work. Log whether it gets it right. This is gate one and gate two, measured.
- Brand mentions, linked and unlinked. An unlinked mention still feeds the corpus. Old SEO ignored these. GEO counts them.
- Sentiment and framing. Being described as "a Raleigh relocation specialist" is a different business outcome than being described as "an agent who also does mortgages sometimes."
- Branded search and direct traffic. When AI visibility works, people search your name and type your URL directly. That lift is the downstream proof.
- Search Console's generative AI performance report. New Google rolled out a generative AI performance report in Search Console in 2026 and has been expanding it to more sites. AI Mode queries also show up in Search Console, tracked separately. Check it monthly next to your inclusion rate.
- Google Business Profile actions. New Calls, direction requests, website clicks, and profile views. Google is also bringing back post insights, so you can see who is actually reading your weekly posts.
Do this today
Open a private window, turn memory off, and ask an AI tool who it recommends for your service in your market. Do not use your own account, because a tool that already knows you will flatter you. Whatever comes back is your actual starting position.
The 30-day version Updated
What should a Realtor or loan officer do in the first 30 days of GEO?
Ten moves, in order, that close the gap fastest. You do not need a six-figure budget or an agency retainer with a twelve-month ramp. You need these done in this sequence, because each one makes the next one work harder.
- Run the three-gate test on yourself. Clean session, three prompts: who are you, where do you work, who should I hire in your market. Write down exactly what comes back. That is your baseline and it will probably sting.
- Fix your name consistency everywhere. Same name, same title, same market, same phone, same specialty across your site, Google Business Profile, brokerage bio, LinkedIn, Zillow, Realtor.com, NMLS, Foursquare, Apple Business Connect, Yelp, and every directory. Contradictions are what make AI guess.
- Kill or update the zombie pages. Old team rosters, previous brokerage bios, dead landing pages, expired profiles. If you cannot get them removed, outweigh them with newer, stronger, owned content.
- Add Person and RealEstateAgent schema to your site, and make sure AI can read it. Or FinancialService plus Person for mortgage. Then run the View Page Source test and check your robots.txt. This is a one-time technical fix with an outsized effect on gate one.
- Rewrite your five most important pages answer-first. Home, about, buyers, sellers, and your top market page. A direct 40 to 60 word answer at the top of each one.
- Add a real FAQ section with FAQPage schema to those same five pages. Use the actual questions clients ask you on the phone, in their words, not keyword-tool phrasing. The rich result is gone. The value to AI is not.
- Rebuild your Google Business Profile properly. Correct primary category, service-area business setup rather than "online only," a full service list with real descriptions, service areas, weekly posts, and photos with real file names. Google's own understanding of you feeds Google's AI answers. Start with the Google Visibility Sprint if you want it done in one pass.
- Ask your last ten clients for a specific review. Give them the prompt: the city, the type of transaction, and what problem you solved. Specific reviews are placement signals. Generic ones are decoration. Never offer anything in exchange.
- Publish one piece of original local proof, built as an answer page. Your own market data, a real client story with permission, a neighborhood breakdown only someone who works there could write. Use the answer page structure above. One piece. Then schedule the next one.
- Get into two places you do not own. A local podcast, a partner page, an association directory, a credible local "best of" list, a news quote. Community threads still help as corroboration. They are no longer a shortcut. Track 2 work starts here and it is where almost nobody bothers to go.
If you only do three
Number one, number two, and number four. Identity and consistency before content. Publishing more blogs while AI still thinks you work in a different state is expensive noise.
Keep going
Related guides and services
- Google + AI Visibility Audit
- AI Operating System Implementation Sprint
- How to Get Found on Google as a Realtor
- AI Visibility for Loan Officers: Raleigh Research
- The Market Authority Engine
- Realtor Authority Website System
- The Visibility Foundation
- Fractional Marketing Partner Support
FAQ Updated
Frequently asked questions about GEO for Realtors and loan officers
What is GEO (Generative Engine Optimization) for real estate?
GEO is the practice of structuring your content, website code, profiles, and third-party proof so AI search tools can identify you correctly, connect you to the market you serve, and cite you inside the answers they generate. For a Realtor or loan officer, it is the difference between being findable and being recommended. SEO competes for a ranking position. GEO competes for a mention inside an answer that most people never scroll past.
Is traditional SEO dead for real estate agents in 2026?
No, and anyone telling you that is selling something. Ranking is still the entry fee for Google's AI Overviews. What changed is how much ranking alone buys you. Ahrefs found the share of AI Overview citations coming from top-ten pages fell from 76 percent in July 2025 to 38 percent by March 2026, because Google now answers the related follow-up questions too. SEO still matters. Covering the whole topic is what turns a ranking into a citation.
How do I get ChatGPT to recommend me as a real estate agent?
There is no submission form. For local hiring questions, a 2026 study of 334 ChatGPT answers found 98.1 percent of cited sources were the business's own website or Google Business Profile, and a median of six agents were named per city. So start with those two, make sure they say the same thing, then back them up with specific reviews, consistent profiles on Foursquare, Apple Business Connect, Yelp, and your brokerage bio, and original local content only you could write.
Why does AI say the wrong thing about my business?
Because your sources disagree and AI resolved the conflict without asking you. Old brokerage bios, outdated team pages, multi-state license listings, a corporate umbrella site, a local data listing you never claimed, and a thin personal presence all compete for the answer. Whichever set of sources has the most weight wins. The fix is removing or outweighing the outdated signals and strengthening the ones you own.
Does FAQ schema still matter now that Google removed FAQ rich results?
Yes. On May 7, 2026, Google stopped showing FAQ rich results in search, and it later dropped support from the Rich Results Test. It did not stop reading the markup. Google still parses FAQ structured data to understand a page, and FAQPage markup still hands AI retrieval systems clean question and answer pairs. Keep it, and make sure the markup matches the visible questions word for word.
How do I measure GEO results when I am not getting clicks?
Replace click-first metrics with visibility metrics. Track inclusion rate, meaning how often you are named when you run your target questions in clean AI sessions. Track entity accuracy, meaning whether AI describes who you are and where you work correctly. Track linked and unlinked brand mentions, sentiment, and downstream lifts in branded search and direct traffic. Then add Google Search Console's generative AI performance report, which rolled out in 2026 and shows how your site appears in Google's AI features.
Can a solo agent or single loan officer compete with a big brokerage here?
Yes, and more easily than in traditional SEO. GEO rewards first-hand experience, original local data, specific reviews, and genuine community presence. In the 2026 study of ChatGPT answers, 73.9 percent of shortlist slots went to named individual agents or teams rather than brokerage brands. In our 2026 Raleigh benchmark only 22 of 75 brands appeared at all in open discovery, which means most local markets are not competitive yet. They are empty.
Should I use AI to write my real estate content?
Use it to speed up outlines, first drafts, and repurposing. Do not publish it unedited. Generic AI-written content fails the credibility filter that AI search itself applies, which is a real irony worth sitting with. A human expert has to add the specifics, the local knowledge, the first-hand experience, and the point of view before it goes live. AI is the assistant. Your experience is the product.
How should a Realtor or loan officer structure a page so AI cites it?
Build each page around one real client question. Put the question in the H1, answer it in 40 to 60 words directly underneath, cover the follow-up questions in their own H2 sections, add local proof only you have, and finish with five to eight FAQs. Link the page up to your service page and sideways to related questions, and mark it up with Article, FAQPage, Person, and BreadcrumbList schema.
Does posting on Reddit still help me show up in ChatGPT?
It helps less than it did, and it was never a foundation. Around August 20, 2026, Reddit citations inside ChatGPT search dropped 86 percent in about four days, and Google has said Reddit gets no special ranking preference. Genuine, helpful participation in local threads can still corroborate who you are. Treat it as supporting evidence, not a strategy you rent.
Do I need an llms.txt file for AI search?
No. Ahrefs looked at 137,210 domains in May 2026 and found 97 percent of llms.txt files were never fetched by anything, and AI bots never went looking for ones that did not exist. It is harmless to have one. It is not a visibility strategy. Clean HTML, clear answers, and consistent profiles do the work.
Does ChatGPT use Bing to find real estate agents?
Less than it used to. Reporting from September 2026 shows ChatGPT now runs its own search index and pulls from several providers, with Bing appearing mainly in deep research. Bing visibility is no longer a reliable stand-in for ChatGPT visibility. Local data sources like Foursquare, your Google Business Profile, and your own website matter more for local questions.
Where should a Realtor or loan officer start with all of this?
Start with a diagnosis, not more content. Run the three-gate test on yourself in a clean AI session, then fix name and market consistency across every profile before you publish anything new. If you want the professional version, a Google + AI Visibility Audit runs your name, website, profiles, reviews, and market signals through the same framework used in the RECSC Raleigh benchmark and hands back a prioritized fix list.
Next step
Find out what AI is currently saying about you.
The three-gate test tells you whether there is a problem. The Google + AI Visibility Audit tells you exactly what is causing it and what to fix first: your entity signals, your website structure, your schema, your profiles, your reviews, and the market signals that decide whether buyers, sellers, borrowers, and referral partners ever see your name inside an answer.
Clear findings. A prioritized fix list. The audit fee credits toward any larger engagement.
Sources for the September 2026 update
- RECSC 2026 Raleigh AI Visibility Benchmark, 75 Raleigh-area real estate and mortgage brands.
- Omni Eclipse, 2026 Real Estate AI Search Report, 334 ChatGPT responses across 197 markets, April to May 2026.
- HousingWire, If ChatGPT picks three agents, how do you get on the list?, August 31, 2026, citing Promptwatch data and the FTC review rule.
- HousingWire, The new real estate playbook is getting cited by AI, not clicked on, June 29, 2026, citing SparkToro and Similarweb.
- Search Engine Journal, Google AI Overview citations from top-ranking pages drop sharply, March 2, 2026, reporting Ahrefs research.
- Ahrefs, llms.txt study of 137,210 domains, May 2026.
- Peec AI, ChatGPT built its own search index, September 2026.
- Real Estate News, What AI reads before it recommends a real estate agent, September 2, 2026.
- Search Engine Roundtable, September 2026 Google Webmaster Report, and AI Overview local results quoting listicles, August 2026.
- Google Search Central, Search generative AI performance reports in Search Console, June 2026.
- The HOTH, Google FAQ rich results deprecated, on the May 7, 2026 change.
About the author. Founder and CEO of Real Estate Concierge Services Co, LLC. Emily Wyatt is a fractional marketing partner for real estate agents, teams, brokerages and mortgage professionals in Raleigh, the Triangle, and nationwide. Real Estate Concierge Services Co builds visibility systems that help people find you on Google, Maps, and AI search. Then they turn that attention into steady lead flow.
RECSC uses strategic content, CRM workflows, AI operating systems, and clear follow-up to help agents and mortgage professionals scale their businesses. If you want marketing that sounds like a human and performs like a machine, start here.
