A buyer relocating for a new job used to start with a Realtor.com search and a spreadsheet. Now they're more likely to open ChatGPT and type something like "best neighborhoods in Austin for a family with two kids under a $650k budget, walkable, good schools." No login, no map filters, no agent yet. Just a conversation that produces a shortlist before a single human is involved.
That shift changes who gets found. A brokerage or agent who ranks on page one of Google for "[city] real estate agent" can still be functionally invisible in that conversation, because the model isn't reading rankings, it's assembling an answer from whatever sources it trusts enough to cite: your Google Business Profile, your MLS/IDX listing data, your reviews, and your published content about the actual neighborhoods people are asking about. This guide covers how that works and what to build so you show up in it.
How real estate buyers and renters are actually using AI search
The scale of the shift is bigger than most agents assume. In a Q1 2026 survey of 4,180 home buyers across 38 U.S. metros, FlyDragon's State of AI Search in Real Estate report found 67% used an AI tool such as ChatGPT, Perplexity, Gemini, Claude or Google's AI Overviews as their primary research method before ever contacting an agent, up from just 17% eighteen months earlier. NerdWallet's 2026 Home Buyer Report separately found 48% of Americans planning to buy in the next year intend to use AI tools somewhere in the process, mostly to estimate costs and understand the steps involved.
I'd treat the precise percentages from any single vendor survey as directional rather than gospel, since none of these are run by a neutral third party with a huge published sample. But the direction is consistent across every source I've checked, including Inman's reporting on Realtor.com building a direct ChatGPT integration so buyers can pull live listing data into a conversation without leaving the chat window. When a portal that big builds for the conversational surface instead of just SEO, that's a signal worth paying attention to.
What's changed isn't just where research happens. It's the shape of the question. Google search trained buyers to type fragments: "3 bed houses Denver." AI chat lets them type the actual constraint they're solving for: budget, commute, school district, walkability, whether the HOA allows short-term rentals. The answer that gets generated pulls from whichever sources cover that specific combination well, which is exactly the kind of narrow, local, factual content most brokerage websites don't have.
What AI models actually pull from when someone asks about your market
Unlike a blue-link search result, an AI answer about "best real estate agent in [suburb]" or "is [neighborhood] good for families" is synthesized from a handful of sources the model trusts. For local, transactional real estate queries, four source types dominate.
- Local business signals: your Google Business Profile listing, review count and star rating, and how consistent your name, address and phone number are across directories. This is the same signal set that drives the traditional Google Local Pack, and AI Overviews and AI Mode lean on the same local index.
- MLS and IDX listing feeds: structured, syndicated listing data (price, beds, baths, square footage, days on market) that flows through your local MLS into IDX-powered pages, and into aggregators like Zillow, Redfin and Realtor.com, which AI models cite heavily for anything property-specific.
- Third-party review and reputation sites: Zillow agent profiles, Realtor.com reviews, Yelp, and niche sites, which models use to answer "best agent for [X]" questions since they aggregate sentiment the model can summarize.
- Published local content: neighborhood guides, market reports and buyer/seller FAQs that actually answer the long, specific questions people ask, which is the one category a brokerage fully controls and most brokerages barely invest in.
The practical implication: you can't optimize for AI search the way you'd optimize a single landing page for a keyword. You're optimizing an entity, meaning the consistent, verifiable footprint of your brokerage or your name across every one of those four source types. Get one of them wrong, like a phone number that doesn't match between your website and your GBP listing, and you create ambiguity a model will often just route around by citing a competitor instead.
Local entity and citation strategy: the foundation
Google has been explicit for years that its local ranking runs on three factors: relevance, distance and prominence. AI Overviews and AI Mode both draw from that same local index rather than building a separate one, so the work here isn't new, it's just higher stakes now because a missing or thin profile doesn't just cost you a map pack position, it costs you the citation entirely.
Google Business Profile: the highest-leverage single asset
According to Whitespark's Local Search Ranking Factors survey, GBP signals carry more weight in local ranking than review signals, on-page SEO and backlinks combined. For a solo agent or a small brokerage, that means your profile is worth more optimization time than your entire website. The fields most agents leave blank, and that move the needle most, are the services and products lists, business attributes, the full business description, secondary categories, and the Q&A section. Profiles that actually fill in all of it read as "complete" to Google's systems, and completeness itself functions as a ranking signal.
- Claim and verify every office location separately if you're a multi-office brokerage, never one profile covering multiple physical addresses.
- Match your name, address and phone number exactly across your website, GBP, Zillow, Realtor.com and your MLS listing agent record. Inconsistency here is the single most common reason a solid agent gets skipped in local citations.
- Answer your own Q&A section with the specific questions buyers actually ask ("do you work with first-time buyers," "do you cover [suburb]") rather than leaving it for randoms to populate.
- Post weekly with market updates or new listings. GBP posts are indexed content, and stale profiles read as inactive businesses to both Google's algorithm and to a model deciding which sources are current enough to cite.
MLS and IDX data: make sure the machine-readable version is right
Most agent websites pull listings through an IDX feed from the local MLS. That feed is what powers your on-site search, but it's also frequently the same structured data that gets syndicated onward. If your IDX provider isn't outputting clean schema markup on listing pages (price, address, property type, availability), you're relying entirely on Zillow or Realtor.com to be the version of that listing an AI model finds, instead of your own site. Ask your IDX vendor directly whether they output Schema.org RealEstateListing markup on rendered pages, not just in a data feed nobody crawls.
Content formats that actually get cited
The content most brokerages publish (generic "5 tips for first-time buyers" posts) almost never gets pulled into an AI answer, because it doesn't answer anything specific enough to be worth citing over a dozen competing sources saying the same thing. The formats that do get cited share one trait: they answer a narrow, local, factual question better than anything else indexed for that exact query.
| Format | Why it gets cited | Minimum viable version |
|---|---|---|
| Neighborhood guides | Answers hyper-local questions (schools, commute, walkability, HOA rules) no national portal covers at street level | 800-1,200 words per neighborhood, updated twice a year, with specific streets and landmarks named |
| Local market reports | Provides the numeric data (median price, days on market, inventory trend) models pull for "is now a good time to buy in [city]" questions | Monthly, sourced from your MLS's public stats, with the raw numbers stated plainly, not just a chart |
| Buyer/seller process FAQs | Directly matches the conversational, multi-part questions people type into ChatGPT or Gemini | 20-30 real questions your clients actually asked, answered in 2-4 sentences each |
| Agent expertise pages | Establishes the entity behind the content, which models weigh when deciding whose local claims to trust | Bio with verifiable transaction history, specific service areas, and designations, not a stock "passionate about real estate" paragraph |
Market reports are worth calling out specifically because they're the easiest win most brokerages skip. Your local MLS almost certainly publishes public aggregate statistics every month. Turning that into a dated, specific page ("Denver Metro Market Report, August 2026: median price $XXX,XXX, YY days on market") gives a model an exact, current, citable number instead of forcing it to guess or default to a national aggregator. I'd rather have twelve months of consistent, boring market reports on a small brokerage site than one viral blog post, because the model is answering a factual question, not looking for personality.
Structured data: the checklist for real estate specifically
- RealEstateListing schema on every active listing page, with price, address, property type and status kept current (stale "sold" listings still marked active is a real, common problem).
- LocalBusiness or RealEstateAgent schema on your homepage and agent bio pages, tying your name to your service area and license number.
- FAQPage schema wrapping your buyer/seller FAQ content, since this is the exact format both Google AI Overviews and ChatGPT's browsing mode parse cleanly.
- Review schema only if the reviews are genuinely embedded and verifiable, since Google has penalized fabricated or scraped review markup in the past.
A 90-day playbook to actually move on this
If you're a solo agent or a 10-person brokerage without a marketing team, here's the order I'd work through it in, roughly a month per phase.
- Month 1, fix the foundation: audit and match NAP across your website, GBP, Zillow and Realtor.com profiles. Fill in every blank GBP field. Confirm your IDX provider outputs real schema markup, not just a widget.
- Month 2, build the local content base: write neighborhood guides for the 5-10 areas that generate most of your actual business, and publish your first dated market report using public MLS stats.
- Month 3, test and iterate: ask ChatGPT, Perplexity and Google AI Mode the actual questions your buyers ask about your market, see who gets cited, and rewrite the pages that lost to a competitor or a national portal.
How to know if any of this is working
Rankings won't tell you this. You have to actually query the tools. Set a recurring monthly check where you ask ChatGPT, Perplexity, Gemini and Google AI Mode the same 10-15 questions a real buyer in your market would ask, things like "best neighborhoods near [landmark] for a young family" or "who's a good real estate agent in [suburb] for a first-time buyer," and log whether you're mentioned, cited, or absent. That log is your actual visibility data, and it's the same core method behind how we track AI visibility for clients across industries more broadly. If you want a structural sense of how this discipline differs from traditional SEO before you build a full plan, our AEO vs. SEO breakdown covers the mechanics. And if you're deciding whether to DIY this or hire it out, our AEO pricing guide and the general playbook for ranking in Google AI Overviews are both good next reads, since most of what earns a citation in Google's AI Overview also earns one in ChatGPT or Perplexity's answer.
The honest cost-benefit for a small brokerage
None of this requires an enterprise budget. A solo agent can do the GBP cleanup and write two neighborhood guides in a weekend. What it does require is treating your online presence as a single connected entity instead of a website, a Zillow profile and a GBP listing that all say slightly different things. The brokerages that will show up in these conversations in 2027 are the ones treating this as infrastructure now, not the ones waiting for a "real estate AI SEO" tool to do it for them.
If you want a second pair of eyes on where your brokerage or your agents currently stand in AI search results, email me at filipe@aipeekaboo.com or book 30 minutes on Calendly and we'll run the actual queries together.
