Record Reliance, New Rules: How Buyers Find a Real Estate Agent Now

A record-high, record-narrow market

Start with what hasn't changed, because it's the part most likely to get lost in a story about disruption. NAR's 2025 Profile of Home Buyers and Sellers, covering transactions from July 2024 through June 2025, put agent use at 88% for buyers and 91% for sellers, the latter matching the highest share on record. For Sale By Owner sales fell to an all-time low of 5%, and those FSBO homes sold for a median of $360,000 against $425,000 for agent-assisted homes, a gap NAR's own data draws directly, not an inference on top of it.

The same report shows the buyer pool itself getting older and more cash-heavy. First-time buyers fell to 21% of the market, the lowest share since NAR began tracking in 1981, and their median age rose to a record 40. All-cash purchases averaged 26% over the year, also a record, and median down payments hit 19% overall, the highest levels in decades for both first-time buyers (10%) and repeat buyers (23%). None of this is a story about people trusting agents less. If anything, reliance on an agent is about as high as NAR has ever recorded it. The open question is what's happening underneath that stability, in how a buyer or seller actually finds the specific agent they end up hiring.

What the settlement actually changed

The Sitzer/Burnett commission-lawsuit settlement, widely reported at $418 million, reshaped the mechanics of that hiring process starting August 17, 2024, with MLSs given until mid-September to implement the rule changes. Two provisions matter most here. Buyer-broker compensation offers can no longer be published on the MLS, and an agent now needs a signed buyer-broker agreement, with objectively stated, non-open-ended compensation terms, before showing a buyer a home.

The practical effect is that a piece of information that used to live, implicitly, on a public listing feed now has to be established in a private conversation before a buyer ever sees a house with that agent. That's not a claim this evidence set can prove drove any specific behavior change on its own. The NAR FAQ page confirms the mechanism and effective date but doesn't restate the settlement dollar figure itself, and no before-and-after study of agent discovery behavior has been published yet. But it is a structural fact: the portal listing carries less of the value-communication load than it did before August 2024, by design.

Where value communication went instead

If the portal listing carries less of that signal, something else has to carry more of it, and NAR's own 2025 REALTOR Technology Survey offers a hint, even if it can't be read as proof of cause and effect. Agents report social media as their top lead-generating technology at 39%, ahead of CRM at 23% and local MLS at just 17%. That's a channel mix that already leans toward direct relationship and personal visibility rather than the listing feed, and it predates as much as it reflects the settlement's effect.

The same survey found 66% of agents adopting new technology primarily to save time and 64% to enhance the client experience, practical motives rather than a wholesale strategy shift. Only 33% rated AI's business impact so far as moderately positive: agents themselves are not yet reporting that AI tools have changed their day-to-day results in a way most of them would call positive. Whatever displacement of value-communication is happening is showing up in channel mix and lead-gen habits, not yet in agents' own read of AI's impact on their business.

The AI-answer layer is arriving fast, even where real estate hasn't been measured yet

Two general-population findings, neither specific to real estate, establish that a new discovery layer is real and moving quickly. Pew Research's Americans and AI 2026 survey, fielded in February 2026 among 5,119 US adults, found 49% of US adults now use AI chatbots, up from 33% in 2024, and 44% specifically use ChatGPT, more than doubling from 18% in 2023. The leading use case chatbot users report is looking up or searching for information, at 42%.

BrightLocal's Local Consumer Review Survey 2026 adds a sharper edge to that: 97% of consumers read online reviews for local businesses before deciding whether to use them, and 41% now say they always read reviews when searching for one, up from 29% a year earlier. More striking is where they're reading those reviews. Google usage as a review-discovery platform fell to 71%, from 83% a year earlier, while AI tools jumped to 45%, from 6%, making AI tools the third most-used review-discovery platform in BrightLocal's survey. We're holding both BrightLocal findings at an emerging tier rather than established, because they rest on a single vendor's survey instrument, even one that's widely cited as an industry standard. What's genuinely missing from public research, and this is the honest center of this study, is any real estate-specific study of whether a buyer or seller asks an AI assistant to recommend or vet an agent the way BrightLocal has measured that behavior for local business generally. That bridge is unmeasured, not disproven.

AI tools went from 6% to 45% of consumers citing them as a place to find local-business reviews in a single year. No one has yet measured whether that reaches real estate agent selection specifically.

Portal concentration is a single point of dependence

Where browsing does happen, it's lopsided. Similarweb's traffic overview, a third-party estimate rather than either company's own audited figures, put zillow.com at roughly 356.5 million visits over the trailing three months as of June 2026, ranking it first among US real estate marketplace sites with about 96% of that traffic from the US. realtor.com logged roughly 109 million visits over the same window, close to a third of Zillow's estimated volume.

Zillow Group's own disclosed active-user and lead-volume figures were not accessible to us through its investor pages, so the traffic numbers here are directionally useful estimates, not the operator's own reporting. Still, the concentration is worth naming: if most portal browsing funnels through one dominant site, a visibility strategy built entirely around a profile on that one platform inherits its concentration risk, especially now that the settlement has pushed part of the value-communication job off the portal entirely.

A wrinkle worth naming before you quote either number

There's a genuine methodological divergence sitting inside NAR's own data that deserves a plain caveat rather than a pick-one-and-move-on treatment. The annual Profile of Home Buyers and Sellers, published November 2025, puts first-time buyers at 21% of the market for the full twelve-month window. NAR's monthly REALTORS Confidence Index for June 2026 puts the same figure at 33% of monthly transactions, alongside cash sales at 25% (versus 26% in the annual figure), 5% of buyers purchasing based on a virtual tour alone with no in-person visit, and 3% of sellers selling to an iBuyer.

The most likely explanation is a difference in sample populations, agent-reported monthly transactions in one instrument versus a full-year buyer survey in the other, rather than a real reversal in buyer behavior between November and June. Both figures are real NAR data. The point for a business owner citing either one is simple: know which instrument you're quoting and say so, rather than presenting either as the single settled number.

What's genuinely unmeasured, and why that's the opening

The single most load-bearing gap in this evidence set is also the most honest one to name directly: no public, retrievable study currently breaks down exactly how buyers found the specific agent they hired, referral from a friend or family member, a prior transaction, an open house, or internet search. The current NAR Profile edition that would carry that breakdown sits behind a member paywall, and no publicly accessible summary reproduces it.

Combine that with the absence of any real estate-specific AI-chatbot-as-agent-discovery study, the absence of Zillow's own disclosed traffic and lead figures, and the absence of retrievable peer-reviewed literature on agent-choice heuristics, and the honest picture is this: the surrounding conditions are well established, agent-mediated transactions at record highs, a settlement that pulled a signal off the portal, and a general AI-answer layer rising fast, but the specific bridge between those conditions and how a buyer or seller actually discovers and chooses an agent is not yet measured anywhere in public research. That's not a weakness in this study so much as the actual frontier.

The surrounding conditions are well established. The specific bridge between them, how a buyer actually finds the agent they hire, is not yet measured anywhere in public research.

Where this points: mapping your own market's attention

National averages tell you the conditions a brokerage or agent is operating inside, not what's true in your specific metro, price tier, or client base. Agent reliance is near a record high nationally. The compensation signal that used to live on the portal listing is gone as of August 2024. A fast-rising AI-answer discovery layer exists in the general population with no real-estate-specific census yet. None of those three facts, by themselves, tell you whether your own footprint today is being found through referral, a portal profile, a review platform, local search, or what an AI engine says when someone asks it to recommend an agent in your area.

That's the gap a Visibility Corpus read is built to close: mapping where a specific market's attention actually sits today, across all of those surfaces at once, rather than assuming the portal or the MLS is still doing the work it used to do. The measurement comes first, in your market, and the answer follows from what it shows.