Eighty-two percent of real estate agents now use AI every day. Their buyers do too. When a first-time buyer in Austin opens ChatGPT and asks who they should hire, they get an answer. The answer is rarely the agent typing prompts into the same model an hour earlier. A new study from 5WPR and Haute Residence put a number on it last month: real estate triggers AI Overviews at a 0.14 percent rate. Health gets 13 percent. Finance gets 4.2. The gap is real, it is measured, and it is fixable inside thirty days.
The data nobody wants to hear
The 2026 Luxury Real Estate AI Discovery Report, published April 23, audited how often ChatGPT, Claude, Perplexity, and Google AI Overviews surfaced real estate professionals in response to buyer- and seller-stage queries. Real estate finished last. Not last among service industries. Last among every major industry the auditors measured. Health pulled 13 percent. Finance hit 4.2. Retail came in at 2.1. Real estate posted 0.14 percent.
The supply side of this story matters more than the headline. Between October 2025 and March 2026, every major American portal shipped or upgraded an AI-powered search surface. Zillow, Redfin, Realtor.com. FBS pushed a server in late April that connects Flexmls accounts to ChatGPT, Claude, and Gemini using existing MLS credentials. The portals are showing up in AI answers. Individual agents are not.
There is a window. The report calls it twenty-four months. That feels generous. Once the brokerage SaaS vendors start shipping AI-visibility audits as a default feature (Luxury Presence already started, with a fresh $37 million from Bessemer), the easy advantage gets priced into every agent’s monthly fee. The agents who move first claim citations the models trust later.
The good news for solo agents and small teams: AI search visibility is mostly an information architecture problem, not a marketing budget problem. The fix is technical content, schema markup, and a short list of external citations. None of it requires hiring a PR firm. You can run the audit this afternoon and start the fix tomorrow.
Run the audit in ten minutes
Open three tabs: ChatGPT (web search enabled), Claude with web tool, and Perplexity. Drop these five queries into each one, substituting your market and niche:
- “Best [city] real estate agent for first-time buyers in 2026”
- “Who should I hire to sell my [neighborhood] home”
- “Top luxury real estate agent in [city]”
- “Real estate agent specializing in [niche, e.g. relocation, new construction, condos] in [city]”
- “[Your name] real estate reviews”
Three questions for every answer the model returns. Do you appear? Who does? What sources did the model cite at the bottom? You will see a pattern. The citations almost always come from Zillow agent profiles, Google Business Profile reviews, third-party “best of” roundups (Expertise.com, ThreeBestRated), brokerage bio pages, and a small number of podcast or local news appearances. The model trusts those sources because the training data already does.
Write down the two or three citation sources that rank competitors in your market. That is your target list for week three. If your name shows up in query five but not the first four, you have a discovery problem, not a reputation problem. Different fix. If your name shows up nowhere, you have a presence problem. Hardest fix, also the most common. Most agents are starting from zero.
The thirty-day fix, week by week
Week one is schema. Add RealEstateAgent schema to your homepage, Person schema to your bio, and LocalBusiness to any office page. Most agents skip this because their site builder does not surface it as an option. Generate the raw JSON-LD, paste it into your site’s head tag (or use a plugin if you’re on WordPress), then run Google’s Rich Results Test before you ship. After it validates, ask Claude or ChatGPT to audit the entity coverage. Did you specify yearsLicensed, areaServed, knowsAbout, makesOffer? Schema is how the model knows what kind of professional you are without guessing.
Week two is page-level content surgery. AI engines do not read pages the way Google’s old crawler did. They extract chunks. Every page on your site should answer a specific question in the first hundred words. Your bio page should not open with a paragraph about your love of golden retrievers. It should open with: “Jane Doe is a real estate agent in Round Rock, Texas, who has closed 187 transactions since 2014, with a focus on first-time buyers and military relocation.” The model can extract that sentence and serve it as an answer. Rewrite your top five pages with the same discipline. Listing pages, neighborhood pages, the bio, the about page, the contact page.
Week three is external citations. The model will not trust your homepage by itself. It cross-references. Fill out your Zillow Premier Agent profile to 100 percent, including specialties, languages, and the long-form “about” field. Do the same on Realtor.com, HomeLight, and your state association profile. Most agents leave half these fields empty. Pitch one local podcast and one community blog this week. You do not need a publicist. A friendly email with three story angles and your contact info still works. Two appearances and three filled-out portal profiles can move your name into citation territory inside a single week.
Week four is the E-E-A-T pass. AI engines pull expertise signals from the same places Google does, plus a few new ones. Add a credentials block to your bio listing license number, year licensed, designations (CRS, ABR, GRI, SRES if you have them), transactions closed, and service area. Add an FAQ section at the bottom of your top three pages, each answering one buyer or seller question in twelve to forty words. Keep it tight. Models prefer short answers they can quote directly.
If you only do one of these weeks, do week three. External citations move the needle faster than schema for most agents, because the schema only helps once the model has decided you are worth investigating. Citations are what trigger the investigation.
A listing description prompt that does not read like AI
Consumer trust in AI to help find a home dropped from 30 percent in 2025 to 16 percent this year, per real estate news coverage from May 9. Buyers can smell generated copy. Most listing descriptions written by ChatGPT default mode trip every signal: “stunning,” “boasts,” “nestled,” “must-see.” A model can sound like a person if you tell it the rules.
Try this prompt with your next listing:
You are writing the listing description for [address].
Hard rules:
- No adjectives without evidence. Do not say "spacious."
Say "32-foot great room."
- No em-dashes.
- Open with the single most unusual feature of this house.
- Close with three practical details: nearest grocery store
by name, commute time to the largest local employer,
monthly HOA cost.
- 180 words, no headings, one paragraph break.
Property facts:
[paste your MLS data sheet]
The output will not be perfect. It will not need to be. It will sound like a person who walked the property, which is the bar buyers are using now that they have read enough AI copy to know the difference.
The window closes faster than the report thinks
The twenty-four-month window in the Haute report assumes brokerage software vendors move at brokerage software speed. They do not always. Luxury Presence shipping AI visibility as a default feature would cut the window in half. Lofty’s Homeowner Agent product, launched in April, already mines existing CRM contacts for likely-seller signals using the same kind of inference real-estate-specific AI is being built around. The agents who pull these four weeks of work forward by ninety days will own citation slots that compound.
Pick your market. Run the five-prompt audit this afternoon. Add schema this week. The hardest part is not the work. It is doing it before everyone else figures out they need to.
If page-level schema and entity markup is the part you would rather hand off, that is a thing we build at AU. If you want the next post in this series delivered every Tuesday morning, subscribe to The AU Brief.
