How do buyers and sellers ask AI about real estate?
Real estate queries to ChatGPT, Gemini and Perplexity cluster into three families. First, agent selection: who is the best listing agent in Riverside, who should I use to sell a condo downtown. The engine names a shortlist and defends it with evidence — review counts, transaction texture, local specialization. Second, process: how much does it cost to sell a house in Austin, how long does probate take in California. Third, area research: is Folsom a good place to buy, which neighborhoods have the best schools near a commute corridor. With Google AI Overviews above 2.5 billion monthly users (Alphabet, Q2 2026) and roughly 68 percent of searches ending zero-click (Ahrefs, February 2026), the agent named inside the answer absorbs the demand that used to be spread across ten blue links.
Google Business Profile, reviews and NAP consistency
Real estate has a structural NAP problem: the agent exists as a GBP entity, a Zillow-class portal profile, a brokerage roster entry and often a personal brand with a different phone number. Pick one canonical identity — agent name, brokerage affiliation, one phone, one service area — and make it byte-identical everywhere. GBP category should be Real Estate Agent or Real Estate Agency; keep it precise. Review operations matter more here than in almost any vertical because the review text is the synthesis material: reviews that mention sold over asking in ten days, guided us as first-time buyers, or knew every street in Willow Glen are exactly what an engine quotes when asked why it recommended you. Aim for a steady cadence, answer every review, and never buy or gate reviews.
Local schema: LocalBusiness, Service and FAQ
Use RealEstateAgent (a LocalBusiness subtype) on your main profile page with name, areaServed, telephone, address and sameAs pointing at your portal profiles and GBP. Mark each service line — buyer representation, listing representation, relocation — with Service schema. Publish your top ten client questions as answer blocks wrapped in FAQPage markup: how do you price a listing, what are your fees, do you work with VA loans. Schema does not guarantee citation, but it removes the engine’s excuse for ignoring facts you already published.
Service-area and neighborhood pages
Agents live and die by hyperlocal relevance. A page per neighborhood you actually work — with real specifics: typical price bands, housing stock, school zones, days-on-market texture, your own recent activity there — is legitimate content and prime citation material for area-research queries. Thin pages with the city name swapped are doorway spam and increasingly filtered. The test: would a resident find the page genuinely more useful than your homepage?
The four-dimension measurement loop
Measure monthly against a fixed question set, under the four-dimension open standard: Visibility (does AI mention you at all), Citation Rate (does it cite your pages as source), Semantic Coverage (which of your question families are answered in your favor) and Brand Perception (0-100 sentiment framing of how it describes you). Log every run with screenshot evidence. The loop: fix facts, publish answers, earn reviews, re-measure. See the measurement standard for the exact protocol.
Pitfalls
- Brokerage brand shadow. If the brokerage owns the reviews and portals, the AI names the brokerage, not you. Build your own entity.
- Outcome promises. Guaranteed sale price claims get hedged or dropped — and invite regulatory trouble.
- Stale listing data. Engines read portals; a profile showing your last sale in 2023 suppresses recommendations.
First-month checklist
- Canonical NAP locked across GBP, portals, brokerage roster and site.
- RealEstateAgent + Service + FAQ schema deployed.
- Ten client-question answer blocks live.
- Three neighborhood pages with genuine local detail.
- Review cadence at 4-6 per month, all answered.
- Baseline four-dimension measurement run and archived.