The industry has spent years scrutinizing Zillow’s lead business, and rightly so. Premier Agent’s ZIP-code auction and the “Contact Agent” routing that sends buyers to agent advertisers rather than listing agents are well documented, and they are now the subject of active litigation in Seattle.
A class-action filed in September 2025 by Alucard Taylor — represented by Hagens Berman, the firm behind the Moehrl commission case — alleges Zillow deceives consumers about who they are actually contacting and conceals the referral fees Flex agents pay.
The case has since broadened: an amended complaint in November 2025 added RICO claims.
That fight will play out in court. But it has crowded out a quieter issue that deserves its own scrutiny — one that touches far more consumers than the contact button ever will: the agent directory itself.
What consumers think the directory is
When a buyer or seller uses Zillow’s “Find an Agent” search, they believe they are looking at an objective leaderboard. They enter a neighborhood, see agents ranked with star ratings and sales figures, and reasonably assume the agent at the top is the strongest performer in that market. The interface presents as neutral. Most consumers have no reason to think otherwise.
It is worth being precise here, because the reality is more nuanced than the lead-routing critique — and the nuance is what makes it persuasive.
What the directory actually measures
Unlike the contact button, the directory is not a straight dollar auction. According to Zillow’s own published methodology, “Find an Agent” surfaces agents based on star rating, review volume, and the sales data agents attach to their profiles via their MLS IDs. Non-paying agents are included; profiles are built in part from MLS records and consumer reviews whether or not the agent advertises.
But the ranking signals are, almost without exception, measures of platform engagement rather than verified market production. Review counts reward the agents who most aggressively solicit reviews on Zillow specifically. Profile completeness rewards the agents who treat the platform as a marketing channel. Self-reported sales reward the agents who diligently feed every transaction back into Zillow’s system. Zillow does not publish the full weighting of its ranking algorithm.
The practical result: an agent’s position in the directory correlates strongly with how much they invest in Zillow as a channel — an investment most closely associated with paying Premier Agents — and only loosely with independently verifiable performance. A genuinely top-producing agent who built a referral-based business and never optimized a Zillow presence can rank below a lower-volume agent who works the platform relentlessly. The directory is better understood as a ranking of platform participation than of professional excellence.
This is not, to be clear, an allegation of fraud. Reviews and sales data are real signals, and many excellent agents rank well. The point is narrower and more important: a resource consumers treat as objective is in fact shaped by who engages with — and pays into — the platform, and that distinction is invisible to the people relying on it.
It is not unique to Zillow
Realtor.com — the portal operated under license to the National Association of Realtors — runs a closely comparable structure: it sells agent leads by ZIP code through Connections Plus and presents a “Find a Realtor” directory built on reviews and self-reported transaction history.
Homes.com markets a more agent-friendly “Your Listing, Your Lead” model that does not divert a listing’s leads to competitors — a meaningful difference — but its visibility is still tiered by spend, with paid members sorting above non-members in search and on neighborhood pages.
The agent-matching services vary the bias rather than removing it. HomeLight and Redfin’s partner program are arguably the closest to merit-based: HomeLight is invitation-only and matches on MLS production data (closed volume, days on market, list-to-sale ratio), and Redfin vets partners on close-rate standards. Yet, both still gate participation behind a referral fee — roughly a third of commission at HomeLight, 30–35% of the buyer-side commission at Redfin — so a high-performing agent who declines to participate is simply absent. Others, such as UpNest, optimize for the agent willing to discount commission most aggressively, surfacing the cheapest agent rather than the most qualified.
The common denominator is the part consumers never see: Across every major consumer-facing platform, no model ranks the genuinely best-performing agent independent of whether that agent pays or opts in. What surfaces at the top is substantially a function of commercial relationships, not independent merit — and there is no neutral, performance-only directory operating at consumer scale for anyone, human or machine, to consult instead.
The new wrinkle: AI is inheriting the bias, not correcting it
There was an assumption that AI-assisted search might finally give consumers an objective read — a tool capable of analyzing real performance data and identifying the genuinely top agents in a market.
So far, the opposite is happening.
When a consumer asks an AI assistant, or a generative search result, “who is the top listing agent in this neighborhood” — or simply “recommend me a real estate agent” — the response carries the tone of independent research. It is not. The model is not querying MLS production records, auditing closings or verifying performance.
For residential agent referrals, it is drawing its source material directly from the portals that dominate the open web and are most readily machine-readable — Zillow and Homes.com chief among them. In other words, the same pay-to-play platforms are functioning as the underlying research library for the AI’s “objective” recommendation. The consumer never sees that handoff.
The bias is therefore not filtered out. It is repackaged. Engagement-and-advertising-weighted rankings become the raw material for an answer that consumers perceive as neutral and authoritative — arguably more trustworthy than the underlying page, because it arrives stripped of the visual cues that might prompt skepticism. The model has not done the underlying work; it has restated, with added confidence, a ranking the portals were paid to shape.
For an industry already wrestling with consumer trust and the post-settlement value-of-an-agent conversation, this is a meaningful development. The mechanism by which the public identifies “the best agent” is being abstracted one layer further from verifiable reality, and made harder to interrogate in the process.
Why this should matter to the industry
For agents, the takeaway is a discipline this profession has always rewarded and a principle we coach relentlessly: never build a business on infrastructure you don’t control. Profiles, reviews and rankings hosted inside a single advertising platform are leased assets. The terms can change — they have changed before — and now an additional layer of automated distribution sits on top, amplifying whatever those platforms decide to surface.
Owned assets — a client database, a genuine brand, a verifiable community reputation, word-of-mouth referral flow — are the only foundation a downstream algorithm change cannot erase.
There is also a constructive response available, and it is not “buy more portal placement.” Because generative engines assemble answers from whatever is public, credible, and machine-readable, the rational play for an agent is to make the verifiable, self-controlled record of their production the easiest thing for those engines to find and corroborate.
n practice that means a fully built Google Business Profile; an owned website — on the agent’s own domain, not a brokerage subpage — that states real production data (homes sold, neighborhoods, price ranges) in plain text and carries structured data markup so it can be parsed:
- Neighborhood-level content that directly answers the questions buyers and AI actually ask
- Consistent name, credentials, and statistics across every public profile so the model can confidently attribute them
- Reviews diversified beyond any single portal, beginning with Google
- Corroborating mentions on independent, credible third-party sources, since models weight what is confirmed across many sites over any one claim.
None of this is a portal subscription. It is the same discipline of owning your distribution, translated for an era in which the first impression is increasingly rendered by a machine. Agents can sanity-check their footprint directly by querying the major AI assistants for the best agent in their market and noting which sources are cited — then closing the gaps those citations reveal.
For the portals and the AI platforms drawing on them, the harder question is one of disclosure. A consumer told they are seeing “the top agents” — whether by a directory or by a chatbot citing one — is entitled to know whether that ranking reflects performance or participation. Right now, most don’t know, and the systems aren’t telling them.
The lead-routing debate will be settled by the courts. The directory question — and the AI layer now amplifying it — is one the industry would do well to confront on its own, before consumers and regulators do it for us.
Tim and Julie Harris are co-founders of Tim & Julie Harris Real Estate Coaching, bestselling authors of HARRIS Rules and hosts of Real Estate Coaching Radio. For daily news, analysis and strategies for real estate professionals, visit Harris Real Estate Daily.The views expressed here are their own.
This column does not necessarily reflect the opinion of HousingWire’s editorial department and its owners.
To contact the editor responsible for this piece: tracey@hwmedia.com
