Real Estate AI Policy: Why Your Brokerage’s Isn’t Enough

By Emily Terrell — Top Coach and Speaker at Tom Ferry International. Speaker for NAHREP, eXp Con, and brokerages nationwide.

A real estate AI policy is your own one-page document governing how you use AI — and your brokerage’s version doesn’t replace it. Under TREC Rule 535.155 and the Fair Housing Act, the license holder owns the advertisement regardless of who drafted it. This guide covers the four sections you need and the tool inventory most agents skip.

Key Takeaways

• Your brokerage’s AI policy protects the brokerage. Yours protects your license. They’re different documents doing different jobs.
• Four sections cover almost all of your real exposure: a review gate, an input rule, a Fair Housing scan, and a rule for generated voice and media.
• The failure mode isn’t a reckless agent. It’s a tool that turned on an AI feature underneath you — which is why the tool inventory matters more than the policy language.
• Fair Housing doesn’t ask who wrote the listing description. It asks whether the language signals a preference based on protected class.
• A policy you write in an afternoon and actually follow beats a twelve-page document you skim once.

What is a real estate AI policy?

A real estate AI policy is a written standard defining which AI tools you use, what information may go into them, what human review happens before AI-assisted output reaches a client or the public, and how AI-generated voice, images, and video get handled. At the brokerage level it’s a supervisory document. At the individual agent level it’s a workflow document — a set of gates built into how you already work.

Most agents assume it’s the brokerage’s job. That assumption is the problem.

Why this matters for real estate agents

Adoption already happened. According to NAR’s 2025 Technology Survey, drawn from a random sample of 49,233 active Realtors, 68% of agents use AI tools in their work. A February 2026 survey of 225 NAR members conducted by Realtors Property Resource put the figure at 82%, with 92% either using AI now or planning to. That same RPR survey found 63% of respondents named accuracy of outputs as their top concern.

Brokerage leadership caught up fast. According to the Delta Media Group Real Estate AI & Leadership Survey reported by HousingWire in 2026, only 1.9% of surveyed brokerage leaders had no plans to adopt AI in 2026 — down from 10.6% in 2024.

Here’s the thing nobody wants to tell you: near-universal adoption plus near-universal uncertainty about accuracy is exactly the condition where compliance problems surface. And they surface on the license holder.

TREC Rule 535.155 defines an advertisement as any communication by or on behalf of a license holder designed to attract the public to use real estate brokerage services — including email, text messages, social media, and the internet. Nothing in that rule cares whether a human or a model produced the words. Responsibility runs in both directions: your name is on the ad, and your broker is separately required to ensure your advertising complies. That’s precisely why one policy can’t do both jobs. Your brokerage’s document manages the brokerage’s supervisory duty. It doesn’t manage your workflow, your tools, or your specific listings.

The National Association of REALTORS® published guidance in March 2026 making the brokerage-level case and offering a customizable AI policy template for brokers. It’s a strong document. It’s also written for the person supervising you, not for you — and the template itself sits behind a member login.

If you’re the one running the training rather than sitting in it, the compliance segment belongs in the room, not in the appendix. I’ve broken down what AI training for real estate agents has to cover separately.

“Your broker’s AI policy is a supervision document. Yours is a workflow document. If the only AI rules in your business live in a PDF you’ve never opened, you don’t have a policy — you have a liability with a cover page.” — Emily Terrell, Tom Ferry Coach

This is general information, not legal advice. Confirm your specific obligations with your broker and an attorney licensed in your state before relying on any of it.

The four sections your own AI policy needs

What outputs require human sign-off before they leave your hands?

Write down which categories of output get reviewed and owned by a licensed human, and which are internal-only. Client-facing goes on the review list: listing descriptions, client emails, market analysis, social captions, video scripts, chatbot replies. Internal-only doesn’t: meeting notes, drafts you’re still working, research summaries you’ll rewrite anyway.

The rule is one line: AI drafts, the agent decides. Everything downstream inherits that. Name a person for each category — usually you, sometimes your transaction coordinator — so the gate has an owner instead of a hope.

What information never goes into an AI tool?

Client names paired with financial details. Loan pre-approval documents. Contract terms tied to an identifiable party. Anything you couldn’t explain to your client if they asked exactly where their data went.

Split your tools into two buckets: business-grade with documented data handling, and everything else. Consumer-tier tools may retain prompts or use them for model improvement. That’s not a reason to avoid them — it’s a reason to know which bucket you’re in before you paste. When I need AI on real client data, I strip identifiers first. The math still works; the exposure doesn’t follow it.

What language triggers a Fair Housing rewrite?

This is the section with the largest downside and the least written-down process across the industry. Build a fixed scan list and run every AI-generated marketing line through it before publishing. Catch these:

• “Perfect for families,” “great for young professionals,” “ideal for retirees” — familial status and age
• “Safe neighborhood,” “quiet, established area” — frequent proxies for race and national origin
• “Walking distance to [church, temple, mosque]” — religion
• School quality used as a neighborhood descriptor rather than a factual amenity
• “Exclusive,” “integrated,” “up-and-coming” — coded community language
• Any sentence describing who the home is for rather than what the home is

Better still, build the constraint into the prompt so you’re preventing the language instead of catching it. Then scan anyway. Generative models trained on decades of listing copy have absorbed decades of listing copy habits.

How do AI voice, images, and video get handled?

The FCC confirmed in a February 2024 declaratory ruling that AI technologies generating human voices fall under the TCPA’s restrictions on artificial or prerecorded voice, which means those calls require the prior express consent of the called party. Your AI-cloned voicemail drop is a robocall in the eyes of the rule. Treat it that way.

For images and video: write down whether generated or substantially altered property imagery gets a disclosure, and where that disclosure appears. Write down whether you use AI avatars of yourself, and whether they’re labeled. If you feature generated people in your marketing, note that consistently generating one demographic creates the same advertising exposure that the words do. The same broker-identification rules that govern your short-form video content apply whether the face on screen is yours or a generated stand-in.

The tool inventory most agents skip

Here’s what I’d actually bet on as the failure mode. Not an agent doing something reckless — an agent using a tool that turned on an AI feature underneath them.

Your CRM adds AI call summaries. Your marketing platform starts auto-generating captions. Your transaction management system introduces an AI document reader. None of that showed up in a decision you made. All of it now touches client data.

So the policy needs a fifth thing, and it’s a list, not a rule: every tool that touches client data, what AI features it runs, and the date you last checked. Review it quarterly. A policy that describes a stack you no longer run is a document about a business you don’t have.

For the deeper build on tool selection and MLS-data handling specifically, see how to integrate AI with MLS systems — most MLS contracts carry their own data-use restrictions that sit on top of everything here.

What about state AI law?

Texas agents ask about this constantly, and there’s a widespread misreading worth correcting.

The Texas Responsible Artificial Intelligence Governance Act (TRAIGA, HB 149) took effect January 1, 2026. It prohibits developing or deploying an AI system with intent to discriminate against a protected class, and the Texas Attorney General holds exclusive enforcement authority with a 60-day cure period. Penalties for curable violations run $10,000 to $12,000; incurable violations run $80,000 to $200,000.

But the affirmative consumer-notice duty — telling people before they interact with an AI system — landed on governmental entities and, in treatment contexts, healthcare providers. The earlier draft would have reached private companies. The enacted version narrowed that out. So no, Texas does not currently require you to disclose AI use to your clients.

That’s the law today, not the law in eighteen months. Build the disclosure capability into your policy now and you won’t be retrofitting it later.

How I use this in my own business

My own policy is one page in the same ops folder as my transaction checklist, and it gets read at the same cadence.

The section I use most is the input rule. On a Stone Oak listing last spring, the seller sent over a full financial picture — payoff figures, a second lien, and the number they needed to clear. I wanted AI help modeling the net-sheet scenarios. What went into the tool was the math with every identifier stripped: no name, no address, no loan numbers. What came back was three clean scenarios I walked her through at the kitchen table.

The review gate is the one my transaction coordinator uses. Nothing client-facing publishes without a licensed human’s name attached to it. That’s not a compliance ritual. That’s the same gate that catches the wrong square footage before a buyer’s agent does.

Common mistakes

Waiting for the brokerage. Your brokerage’s document manages their supervisory duty. It doesn’t manage your listings, your CRM, or your Instagram captions.

Writing twelve pages nobody reads. One page you follow beats a policy manual you skim once during onboarding. Length is not rigor.

Treating the Fair Housing scan as a vibe check. “I’d notice if it were discriminatory” is not a process. A fixed list is a process.

Skipping the tool inventory. The policy language is the easy part. Knowing what’s actually running is the part that expires.

Assuming the vendor absorbed the liability. Read the terms. In almost every case, the tool provides output and you provide judgment — and the license holder provides the license.

Never revisiting it. NAR guidance, state law, and your own tool stack all moved in the last twelve months. Quarterly review, on the calendar, or it doesn’t happen.

Frequently Asked Questions

Does my brokerage need an AI policy?

Yes, and so do you. Your brokerage’s policy addresses their supervisory obligation over advertising, disclosures, and client communications. NAR published broker guidance and a customizable template in March 2026. But that document governs the brokerage’s exposure, not your daily workflow, your tool stack, or the specific listings carrying your name.

What should a real estate AI policy include?

At the individual agent level: a review gate defining which outputs need human sign-off, an input rule defining what client information never enters an AI tool, a Fair Housing scan list run before any marketing publishes, rules for AI-generated voice and media, and a quarterly inventory of every tool touching client data. NAR’s broker-level template adds scope, approved tools, training, and enforcement.

Do real estate agents have to disclose AI use to clients?

No federal rule currently requires blanket disclosure that AI assisted your marketing. Specific uses do carry duties — AI-generated voice calls require prior express consent under the TCPA per the FCC’s February 2024 ruling, and altered or generated property imagery carries accuracy and misrepresentation exposure. Check your state commission and MLS rules, which vary and are changing.

Can I put client information into ChatGPT?

Not into consumer-tier tools without understanding their data handling. Many retain prompts or use them for model improvement. Use business-grade tools with documented data policies for anything involving client PII, or strip identifiers before the data goes in. The test: could you explain to your client exactly where their information went?

Does Texas law require agents to disclose AI use?

Not currently. TRAIGA took effect January 1, 2026, but its consumer-notice requirement applies to governmental entities and, in treatment contexts, healthcare providers — not to private real estate businesses. TRAIGA’s prohibition on deploying AI with intent to discriminate does apply broadly. The Texas Attorney General enforces it exclusively, with a 60-day cure period.

Who is liable if AI writes a Fair Housing violation into my listing?

You are. The Fair Housing Act addresses the advertisement, not the authorship. TREC Rule 535.155 places advertising responsibility on the license holder, and separately requires brokers to ensure sponsored agents’ advertising complies. “The tool wrote it” has no standing in either framework.

How long should my AI policy be?

One page. If it runs longer, you’ve written a document you’ll reference once. The goal is a set of gates you actually follow, not comprehensive coverage of every hypothetical.

How often should I update it?

Quarterly, tied to the tool inventory. The policy language changes slowly. Your tool stack doesn’t — features get added to platforms you already pay for, without a decision on your part.

Bring this to your team or event

Emily Terrell speaks at brokerage events, real estate conferences, and team trainings on AI, systems, and social media — the exact playbook in this post, delivered live to your audience. As a Top Coach and Speaker at Tom Ferry International and an active agent closing 70+ transactions a year, Emily speaks from the stage about what’s working right now, not theory. Recent stages include NAHREP and eXp Con.

Book Emily to speak at your next event:
Email: eterrell@yourcoach.com
Phone: (210) 400-9191
Web: coachemilyterrell.com

For real estate agents who want to implement this: Get the weekly real estate prompt library at weeklyrealestateprompts.com or follow @coachemilyterrell on Instagram for daily systems and AI breakdowns.