In April 2026, FlyDragon published the largest study of AI search behavior in U.S. residential real estate ever conducted. They analyzed 12,400 AI-generated responses, tracked 8.2 million queries across 192 metros, and surveyed 4,180 homebuyers.

The headline number: 67% of homebuyers now use an AI tool — ChatGPT, Perplexity, Google AI Overviews, Gemini, or Claude — as their primary research method before contacting an agent. Eighteen months ago, that number was 17%.

This is the fastest behavioral shift in real estate marketing history. And it's not slowing down — it's accelerating.

But here's where it gets uncomfortable for most agents: only 8.4% of practicing U.S. agents appear in any AI-generated response to high-intent searches in their own market. The other 91.6% are completely invisible to the channel that buyers are moving to fastest.

The Traffic Didn't Go to Another Portal — It Went to AI

For the first time since FlyDragon began tracking in 2024, Zillow's share of agent-discovery traffic declined year-over-year — from 41.2% to 33.8%. That's a 17.5% relative drop in twelve months.

The displaced traffic didn't migrate to Realtor.com, Redfin, or another portal. It moved to AI search tools. This isn't a blip. It's a structural shift in how buyers find agents.

The portal model was built for a buyer who searches in fragments — type a keyword, scan a list, compare headshots, call three agents. That buyer still exists, but a growing share of the market now behaves differently. They have a conversation with an AI, get educated, get a recommendation, and reach out to one agent.

Why AI Leads Close at 4x the Rate of Portal Leads

Across 42,180 tracked leads in the FlyDragon study, the performance gap between AI-sourced leads and portal leads is staggering.

The reason is straightforward. A buyer who has spent 30+ minutes asking an AI about a market, neighborhood, pricing trends, and agent recommendations arrives pre-educated. They speak to agents the way a referral does — with context, intent, and a shorter decision timeline. They've essentially pre-qualified themselves before making contact.

Second, AI tools rarely surface more than 3–5 agents per search. Compare that to hundreds or thousands of agents competing for the same Zillow lead. The competition at the point of discovery is dramatically lower, which is why the close rate is dramatically higher.

The Winner-Takes-Most Distribution

The FlyDragon data reveals a concentration pattern that should alarm every agent not actively working on AI visibility: the top 1% of visible agents capture 47% of all AI citation share across metros.

This isn't like SEO where positions 1 through 10 all get some traffic. AI search is winner-takes-most. The model recommends 2–3 agents. If you're not one of them, you get zero from that query — not less, zero.

The flip side is equally important: in 71% of U.S. metros, no single agent currently holds more than 15% citation share. The dominant AI position is unclaimed in three out of four American markets. But the compounding nature of AI citation — where being cited leads to more data, which leads to more citations — means the window to claim that position is narrowing every quarter.

Why Most Agents Don't Show Up

The concentration is partly a training-data problem. Zillow, Realtor.com, Redfin, Trulia, and Homes.com collectively account for an estimated 61% of real estate-related URLs in publicly available LLM training datasets. The default frame most AI models use to answer real estate questions is portal-shaped — agents are presented as line items inside portals rather than as independent professionals.

Breaking through that default requires building an identity outside the portal context. The study found that agents with citations spread across four or more review platforms are significantly more likely to surface in AI responses than agents with all reviews concentrated on a single site — even when the single-site agent has a higher total review count.

AI models don't just count reviews. They evaluate consensus across independent sources. An agent with 50 Google reviews, 30 Zillow reviews, 15 Yelp reviews, and 10 Realtor.com reviews signals broader authority than an agent with 200 Zillow reviews and nothing else.

What the Visible 8.4% Are Doing Differently

Based on the HousingWire and FlyDragon data, along with iPullRank's research on how AI search engines decompose queries through "query fan-out," here's what agents who actually appear in AI results have in common.

1. They have their own website with original local content. Not a portal profile, not a brokerage sub-page — a standalone website with neighborhood guides, market data, and content that directly answers the hyper-local questions buyers ask AI. AI models break a single buyer question into dozens of sub-queries. Pages with specific, data-backed, atomic answers are significantly more likely to be cited than generic overview content.

2. They have reviews distributed across 4+ platforms. Google Business Profile, Zillow, Realtor.com, Yelp, Facebook, Nextdoor — the specific platforms matter less than the distribution. AI models triangulate trust from multiple independent sources.

3. They have consistent NAP (Name, Address, Phone) data everywhere. AI models build entity graphs. If your name is "Sarah Johnson" on Google, "Sarah J." on Zillow, and "S. Johnson Realty" on your website, the model may treat these as separate entities rather than one authoritative agent.

4. They have third-party citations. Mentions in local news, industry publications, podcast appearances, guest posts on real estate blogs — these function as "votes of confidence" that AI models weight heavily. A single mention in HousingWire or Inman carries more citation weight than dozens of self-published social media posts.

5. They use structured data markup. Specifically, RealEstateAgent schema on their website that explicitly tells AI models their name, credentials, service area, and specializations in machine-readable format. HousingWire's follow-up reporting confirmed that agents using structured schema are overrepresented in AI citations relative to their market share.

6. They respond fast and have the data trail to prove it. Speed-to-lead data shows that leads contacted within 5 minutes are 21x more likely to be qualified, and 78% of buyers work with the first agent who responds. AI models increasingly factor response-time reputation and review sentiment about responsiveness into their recommendations.

The Economics of Ignoring This

Let's do the math using data from our 2026 Statistics Hub.

A Zillow Premier Agent lead costs $20–$150+ depending on market. At a 2.4% close rate, you need 42 leads to close one deal. At $50/lead average, that's $2,100 in lead cost per closed transaction — before you factor in your time nurturing 41 leads that didn't convert.

An AI-sourced lead, when you've earned the citation, costs effectively $0 in marginal terms and closes at 9.6%. That's roughly 10 leads to close one deal, at no per-lead cost. The GCI per closed AI lead is $1,180 versus $240 for Zillow.

Even if you invest $500/month in content creation, review generation, and structured data optimization to earn AI citations, the cost per closed deal is a fraction of what portal advertising costs. And unlike portal leads, AI visibility compounds over time — every piece of content, every review, and every citation makes the next one more likely.

What to Do This Week

You don't need to overhaul everything at once. Here's what moves the needle fastest based on what the data shows.

Step 1: Run the visibility test. Open ChatGPT, Perplexity, and Google AI Overviews. Search "best real estate agent in [your city]" and "who should I hire to sell my home in [your city]." See if you appear. If you don't, that's your baseline.

Step 2: Claim and optimize your Google Business Profile. This is the single highest-impact action. It drives 33% of all local clicks and is one of the first sources AI models check. Add your full service area, specializations, and request reviews from recent clients this week.

Step 3: Distribute your reviews. If all your reviews are on one platform, start asking clients to leave reviews on a second and third platform. The 4+ platform threshold is where AI citation likelihood jumps significantly.

Step 4: Publish one piece of original local content on your own website. Not a portal blog, not a social media post — a page on your website that answers a specific question a buyer in your market would ask an AI. "What are closing costs in [city] in 2026?" or "Best neighborhoods in [city] for families with school-age kids" — these are the atomic queries AI tools decompose into.

Step 5: Fix your NAP consistency. Google your name in quotes. Check every listing that comes up. Make sure your name, phone number, and brokerage are identical everywhere. One hour of cleanup work now prevents the entity-fragmentation problem that keeps agents invisible.

The Window Is Open — But It's Closing

The data is clear. Buyers are moving to AI search faster than any previous channel shift. The agents who are visible in AI results close deals at 4x the rate of portal leads, at a fraction of the cost. And in three out of four U.S. markets, the dominant position is still unclaimed.

But AI visibility compounds. The agents who start building their citation footprint now will be exponentially harder to displace twelve months from now. The agents who wait will find themselves competing for an ever-shrinking share of portal traffic that costs more every year.

The question isn't whether this shift is happening. The data settles that. The question is whether you'll be one of the 8.4% who are visible — or one of the 91.6% who are not.