Bottom Line: Brokers and agents who submit listings to the MLS cooperative accept a compact. That compact governs what every other participant in the cooperative can and cannot do with that data. Today, that compact is being violated at an industrial scale. AI search platforms including ChatGPT, Gemini, Perplexity, and others are surfacing MLS listing data, explicitly citing Zillow and other portals as the source. The mechanism that makes this possible is a deliberate technology choice by Zillow and others. MLSs must act now to close the policy gap, or they will have surrendered the foundational asset of organized real estate to platforms that owe them nothing.
The Compact Nobody Is Defending
Every broker who submits a listing to an MLS does so under a set of rules. Those rules govern who can display that data, on what platforms, under whose control, and for what purpose. The IDX policy that NAR has administered since 2000 was built on a single governing principle: the display of MLS listing data must remain under the actual and apparent control of a participating broker.
That principle was stress-tested repeatedly over two decades. When franchise brands wanted IDX feeds on national websites, the industry said no — franchise brands are not themselves MLS participants. When brokers wanted to embed IDX on Facebook business pages, the industry negotiated a narrow permission with specific guardrails: broker-controlled pages, no crawlable data, clear attribution. Every expansion of the IDX footprint required deliberation, debate, and policy action by the MLS community.
That deliberative process has now been bypassed entirely. Not by accident. By design.
What Zillow And Others Did — and How They Did It
In October 2025, Zillow became the first real estate company to launch an app inside ChatGPT, allowing consumers to search properties by conversational description and receive live, interactive results with photos and maps directly within the chat interface.
The technical mechanism Zillow used was a Model Context Protocol (MCP) server. MCP is an architecture standard that allows external applications to query a data source and return results inside an AI chat environment. When Zillow connected its MCP server to OpenAI’s Software Development Kits, it effectively republished MLS listings in an environment that no broker, MLS, or participant controls.
Zillow’s defense is that it maintains “100% control over the app experience.” That argument does not survive scrutiny. From the consumer’s perspective, they are using ChatGPT and simply invoking Zillow as a tool inside that environment. The consumer did not navigate to Zillow.com. They did not engage a broker-controlled interface. They interacted with OpenAI’s platform, which OpenAI controls, monetizes, and trains on — and Zillow’s listing data populated that interaction.
The IDX licenses Zillow holds permit display of MLS data on Zillow.com and its mobile applications. They do not authorize Zillow to republish that data on platforms owned and operated by other companies. NAR’s Handbook on Multiple Listing Policy states that IDX listings must be displayed “under the actual and apparent control of the licensed participant” and that participants may not redistribute MLS data to any person or entity not authorized by the MLS.
OpenAI is not authorized by any MLS. It never has been.
The Robots.txt Deception
There is a second layer to this story that has received almost no attention, and it is arguably more troubling than the MCP integration.
Zillow’s public-facing robots.txt file — the machine-readable protocol that instructs web crawlers what they may and may not access — imposes rigid restrictions limiting automated scraping. Critical directives prohibit access to key areas including /homes/ and /homedetails/, the most listing-rich sections of the site. These restrictions apply to general user-agents.
However, the major AI crawlers — GPTBot for OpenAI, Googlebot for Gemini, ClaudeBot for Anthropic, and others — operate under a separate and fundamentally more permissive set of rules. Site owners can grant or deny access to these bots independently. The question that no MLS has yet formally investigated is which AI crawlers has Zillow explicitly permitted to index its listing pages.
This matters because AI companies have frequently ignored robots.txt entirely. A June 2024 Reuters report found that AI companies often ignored robots.txt directives, and a May 2025 prepublication paper from Duke University confirmed that many bots never checked the robots.txt file at all, and that compliance dropped as restrictions became stricter.
On August 4, 2025, Cloudflare published detailed findings accusing Perplexity AI of deploying stealth crawlers — bots that ignore robots.txt, do not identify themselves, and mimic real browsers to evade detection. These bots reportedly used generic browser headers and IP rotation, techniques commonly associated with crawlers attempting not to be seen.
When a consumer opens Gemini or Perplexity and asks about homes for sale in a specific ZIP code, and that AI returns listings attributed to Zillow or other portals, the chain of custody for that data runs directly through web crawling infrastructure that operates entirely outside the MLS cooperative framework. No broker authorized that crawl. No MLS licensed that access. No IDX agreement governs that display.
The Policy Vacuum NAR Left Behind
NAR’s IDX rules date back to the year 2000. They have been tweaked over the years, but they center on what participants are not supposed to do with MLS data, not on how brokers can innovate in the display of data or facilitate new applications.
When Zillow’s ChatGPT integration triggered industry alarm in October 2025, NAR essentially deferred on policing compliance, issuing a statement that each MLS is individually responsible for conducting its own assessment of technologies that use and display MLS data.
That is an abdication. NAR administers the policy framework that defines participation in the MLS cooperative. It sets the rules. It should enforce the standards. Leaving 580-plus MLSs to independently assess the same technology question is not governance — it is organized confusion. Zillow is operating in hundreds of MLS markets simultaneously. The answer to whether Zillow’s AI integration violates IDX policy should not vary by zip code.
The Council of Multiple Listing Services acknowledged that IDX policy was written in a different era and should evolve with technology, not behind it. That is correct. But acknowledgment without action is not leadership.
The Proliferation Risk Is Already Here
What started as one Zillow-ChatGPT integration is now an industry-wide pattern. Redfin launched its own ChatGPT app integration in February 2026, allowing users to search listings and access market data through conversational AI inside ChatGPT. Realtor.com followed in March 2026 with its own ChatGPT integration focused on pre-search conversations and listing previews.
Realtor.com has also launched RealAssist AI, a conversational assistant built on Google’s Gemini that lets buyers describe what they want in plain language and receive listings, pulling in MLS data without the traditional filter interface.
Every one of these integrations was built on MLS data that brokers submitted under cooperative agreements. None of these platforms — not OpenAI, not Google, not Perplexity — hold a data license with any MLS. WAV Group observed in October 2025 that the industry now lacks a clear framework or guidelines for MLSs to approve or deny AI-based IDX displays. We are operating in a policy vacuum.
That vacuum is being filled by the portals. They are making unilateral technology decisions and presenting MLSs with a fait accompli. This is the same playbook Zillow used when it launched the Zestimate, when it launched Flex, and when it unilaterally imposed the Listing Access Standards program. Ask forgiveness later, if anyone can even agree on what the rules say.
What Brokers and MLSs Must Understand
The broker who submits a listing to the MLS does not surrender that listing to the portals to do with as they wish. The cooperative compact creates specific, bounded permissions. A broker’s listing is not Zillow’s to feed into OpenAI. It is not Google’s to train Gemini. It is not Perplexity’s to scrape and redistribute. It belongs — for display purposes — to the licensed participant in the MLS cooperative, under the rules that participant agreed to follow.
When AI search returns a listing from Zillow, the broker whose listing is displayed has not consented to that channel. The seller whose home is featured has not consented. The MLS that licensed the data to Zillow for a specific, bounded purpose has not consented.
This is not a gray area. It is a clear breach of the cooperative compact on which organized real estate is built.
Strategic Recommendation for the Board
MLSs cannot wait for NAR to resolve this. The portals will not wait. Here is what MLS boards and senior executives must do immediately.
Conduct a data audit. Identify every AI platform — ChatGPT, Gemini, Perplexity, Copilot, and others — that is returning your members’ listings in response to consumer queries. Document the source attribution. Preserve the evidence.
Review and update IDX data license agreements. Every IDX agreement should include explicit language prohibiting licensees from sublicensing, transmitting, or republishing MLS data to AI platforms, LLMs, or any third-party domain not explicitly authorized in the license. This language does not exist in most current agreements. It should. I believe that BrightMLS has done this – check with them.
Require AI use disclosure. Any licensee that integrates MLS data into an AI product or platform should be required to disclose that integration to the MLS in advance and demonstrate compliance with the cooperative’s display and control requirements before the integration goes live.
Enforce existing agreements. Where portals have already deployed AI integrations without authorization, MLSs should put them on notice. Data feed suspension is the most powerful enforcement tool MLSs have. They should be willing to use it.
Develop AI-specific data governance policy. NAR’s Multiple Listing Issues and Policies Committee should take up AI data use standards as an emergency agenda item. The deliberative process that has governed IDX policy for 25 years must be applied to AI — not bypassed by it.
The MLS exists because brokers decided that cooperative sharing of listing data would benefit consumers and the industry. That decision was not made to benefit OpenAI, or Google. The MLS governs the most valuable, cleanest, most structured real estate dataset in existence. It is not an open commons for AI platforms to harvest at will.
Stop giving AI another broker’s listing. It was never yours to give.
Wondering whether your MLS or brokerage is unintentionally exposing listing data to AI platforms? WAV Group works with MLSs, brokerages, and industry organizations to assess AI data governance, review licensing policies, and develop practical strategies that protect broker-owned data while supporting innovation. Let’s start the conversation.
