What Should Real Estate Agents Never Use AI For?
By Emily Terrell — Top Coach and Speaker at Tom Ferry International. Active San Antonio agent closing 70+ transactions a year.
Real estate agents should never use AI for three things: cloned-voice outbound calls, generated client testimonials, and drafting or interpreting contract language. Each one converts an AI output into a legally operative act that a license or a federal rule reserves for a human. Everything else on your desk needs a review step, not a prohibition.
Key Takeaways
- Only three AI uses are categorically off-limits — everything else is a review problem, not a ban.
- AI voice cloning on outbound calls falls under the TCPA, and the FCC confirmed it in February 2024.
- AI-generated testimonials and reviews carry civil penalties under the FTC’s rule, effective October 2024.
- Asking AI to draft an escalation clause or explain a contingency is the unauthorized practice of law in Texas, indirectly counts, and your license doesn’t cover it.
- The durable fix isn’t a longer “never” list. It’s one approval gate that three categories of output have to pass through.
What does “never use AI” actually mean in real estate?
“Never use AI for X” is a task list, and task lists don’t survive the next model release. The more useful frame is that there’s almost no task AI can’t touch — but there are moments where an AI output stops being a draft and becomes a legally operative act. Those moments require a human who holds a license, or a fact that actually happened.
That distinction is the whole post. Three uses fail that test no matter how carefully you review them, because the problem isn’t quality. It’s authority. The rest of what agents worry about — listing copy, images, valuations, client data — is real exposure that gets managed with a review step, not a prohibition.
Why this matters for real estate agents
Adoption already happened. The guardrails didn’t.
RPR’s February 2026 survey of 225 NAR members found that 82% of agents currently use AI in their business and 68% use it daily or several times per week. The same survey found compliance or legal issues was the second-most-cited concern at 49%, behind accuracy of outputs at 63%, with Fair Housing concerns at 28% (RPR, February 2026).
Read those two sets of numbers together. Two-thirds of the industry is producing AI output several times a week, and half of them already know they’re not sure where the line is. That gap is not a technology problem. It’s a missing approval gate.
“The three things you can’t delegate to a model aren’t the scary ones. They’re the boring ones — a voice on a phone call, a testimonial about a closing, and a sentence in a contract. Every one of those is a legal act, and a legal act needs a person behind it.” — Emily Terrell, Tom Ferry Coach
The three uses that are actually never
Should real estate agents use AI voice cloning for outbound calls?
No. On February 8, 2024, the FCC adopted a Declaratory Ruling confirming that calls made with AI-generated voices are “artificial” under the Telephone Consumer Protection Act (FCC, February 2024). The ruling took effect immediately.
What that means in practice: an AI voice agent working your expired list, your FSBO list, or a circle-prospecting campaign is making artificial-voice calls under the TCPA. Marketing calls require prior express written consent. Not a phone number you pulled from public record. Not a lead that filled out a form eighteen months ago. Written consent, per the standard the statute sets.
This is the item on the list with the worst math, because TCPA damages accrue per call and AI voice tools are sold on volume. A product whose entire pitch is “it dials two hundred numbers while you’re at a listing appointment” is a product that scales your exposure at exactly the same rate it scales your dials.
The narrow version that works: AI voice on inbound calls you’ve been given permission to answer, or as a transcription and summary layer on calls you’re already making. The moment the AI voice initiates contact, you’re in TCPA territory.
Can agents use AI to write client testimonials or reviews?
No, and this one catches good people who weren’t trying to cheat.
The FTC’s final rule on consumer reviews and testimonials, 16 CFR Part 465, was announced August 14, 2024 and specifically targets reviews and testimonials that misrepresent they came from someone who doesn’t exist — the announcement names AI-generated fake reviews directly — or from someone who didn’t actually have the experience described. The rule allows the agency to seek civil penalties against knowing violators (FTC, August 2024).
Here’s how agents fall into it without meaning to. You had a great closing. The client said something nice in a text but won’t sit down and write a review. So you ask a model to “write a testimonial from a happy seller in Stone Oak,” drop it on a graphic, and post it. The sentiment was real. The testimonial wasn’t. That’s the violation.
The same rule reaches two other things agents do: buying reviews with incentives conditioned on writing something positive, and testimonials from insiders — officers, managers, employees, agents — that don’t disclose the connection. A “client win” post written by your own team member is covered.
What works instead: use AI to format a real testimonial. Take the client’s actual words, ask the model to tighten them for a graphic, send it back to the client, and get their approval on the version you publish. The words originate with a real person who had a real experience. That’s the entire difference.
Can AI draft or explain contract language?
No, and Texas wrote the rule in a way that makes the AI question easy to answer.
Under 22 Texas Administrative Code § 537.11(b), a license holder may not practice law, may not “directly or indirectly offer, give or attempt to give legal advice,” may not give opinions on the legal effect of contract forms, and may not draft or recommend language defining or affecting the rights, obligations, or remedies of the principals — the rule names escalation, appraisal, and contingency clauses specifically (Texas Administrative Code § 537.11).
The word doing the work is indirect. An agent who asks a model to draft an escalation clause and then forwards it to a buyer has indirectly done the thing the rule prohibits. The model isn’t licensed. Neither, for this purpose, are you.
The same applies to interpretation. “Explain to my buyer what this financing contingency means if the appraisal comes in low” is a request for a legal opinion on the effect of a contract term. It doesn’t become something else because a chatbot answered it.
The rule does leave room. Section 537.11(c)(2) preserves your ability to explain the meaning of informational items or choices in a contract form, as long as you’re not practicing law or giving legal advice. And 537.11(d)(2) permits adding or striking language when a principal specifically instructs you in writing, with the change made conspicuous. Agents outside Texas need to check the equivalent rule from their own commission — every state has one, and they’re not identical.
What works: AI as a comprehension tool for you, not an output tool for the client. Ask the model to explain a concept so you understand it better before you refer the client to their attorney. Nothing generated goes into the file or the conversation.
What people think is “never” but is actually “review”
Most of what agents put on their mental never-list belongs in a different bucket. These are real exposures managed by a review step, and I’ve covered each one in depth:
- Listing copy and Fair Housing. Models will produce “perfect for young families” or “safe neighborhood” without hesitation, because that language is statistically normal in the training data. The Fair Housing Act applies to the advertisement regardless of what drafted it (NAR Fair Housing). Build the constraint into the prompt, then read every line before it goes anywhere.
- Altered or generated listing images and video. Disclosure, not prohibition — and in California it’s law. The full three-tier test is in what real estate agents have to disclose when using AI.
- Broker identification in AI-generated captions. The most common Texas violation happening right now, and it’s covered in what actually risks your license.
- Client data in consumer AI tools. Know what the tool retains before anything sensitive goes in. Same post covers the standard.
- AI valuations. Support for a CMA, never the CMA.
This is general information, not legal advice. TCPA, FTC, and state license rules vary and change. Consult your broker and an attorney licensed in your state before setting policy.
How I use this in my own business
I close 70+ transactions a year in San Antonio on roughly five hours a week of active management, and AI touches nearly everything I produce. It touches none of the three items above, and that’s a system, not restraint.
Three categories of output route through an approval gate before anything publishes: property copy, anything that makes a claim about a past client, and anything that touches the contract. Everything else goes straight out. That’s a one-page document, not a philosophy.
The testimonial rule is the one I had to build deliberately, because it’s the one that feels harmless. A past client sent me a text after closing that was better than anything I could have written. I asked her to approve a cleaned-up version for a graphic, and she did — in writing, in the same thread. That thread lives in the transaction file. Ninety seconds of work, and the testimonial is now something I can defend rather than something I hope nobody asks about.
On voice, I made a decision early and haven’t revisited it: no AI voice agent initiates a call in my business. The tools got good fast, and the pitch is genuinely compelling when you’re staring at a list of 300 expireds. The consent math doesn’t work, so the answer is no before I look at the demo.
Common mistakes
Treating the never-list as the compliance system. The list changes every quarter. The approval gate doesn’t. Agents who memorize prohibitions instead of building a gate get caught by the item that wasn’t on the list yet.
Assuming an AI voice vendor solved consent for you. Vendors sell capability. The call is placed on your behalf, to your list, for your business. Ask any vendor to show you exactly how consent is captured and documented per number, and read the answer carefully.
Publishing a testimonial with real sentiment but generated words. The FTC rule reaches the representation, not your intent. If the person didn’t say it, they didn’t say it.
Letting AI answer a client’s contract question because you were in a hurry. Speed is the reason this happens. It’s also why “the AI explained it” reads badly in a complaint.
Confusing “AI drafted it” with “AI is responsible for it.” No regulator that has weighed in has accepted that framing. You own every output the moment it leaves your hands.
Building the list and skipping the routing. Knowing the three items is worthless if nothing in your workflow forces a stop. Write down which three categories require a human sign-off and who signs.
Frequently Asked Questions
Is it illegal for real estate agents to use AI?
No. No state prohibits licensees from using AI, and no federal rule bans it in real estate. Enforcement runs through what the content says and does, not what produced it. The three genuine limits — AI voice on outbound calls, generated testimonials, and contract drafting or interpretation — are restrictions that already existed and now apply to a faster production method.
Can I use an AI voice assistant to call my leads?
Not for outbound marketing calls without prior express written consent. The FCC confirmed in February 2024 that AI-generated voices are “artificial” under the TCPA, which means AI voice calls carry the same consent requirements as any prerecorded marketing call. Inbound answering and call transcription sit on much safer ground. Confirm with your broker and counsel before deploying any voice tool.
Can AI write client testimonials for my marketing?
No. The FTC’s rule on consumer reviews and testimonials, announced August 2024, specifically covers AI-generated testimonials that misrepresent a real person’s experience, and it allows civil penalties against knowing violators. Use AI to tighten a real client’s actual words instead, then get the client’s written approval on the published version. Real experience, real person, documented consent.
Can I ask ChatGPT to write an addendum or explain a contract clause?
No. In Texas, 22 TAC 537.11 prohibits license holders from practicing law, from directly or indirectly giving legal advice, and from drafting language affecting the rights, obligations, or remedies of the parties — escalation and contingency clauses are named specifically. Routing the request through an AI tool doesn’t change the analysis. Refer the client to their attorney.
What can real estate agents safely use AI for?
Most of the work. Drafting listing copy you then verify and read for Fair Housing, first-pass follow-up emails, social captions, market summaries, content calendars, transcription, research, and prep for listing appointments. The common thread is that a licensed human reviews the output before a consumer sees it. AI drafts, the agent decides.
Do I need a written AI policy for my team?
Yes, and it should be short. Name the approved tools, name what client data never goes into them, and name the three categories of output that require a human sign-off before publishing — property copy, past-client claims, and anything touching the contract. A one-page document beats a training session nobody remembers by Thursday. Brokerage leaders should confirm the specifics with counsel.
Does using AI void my E&O coverage?
Ask your carrier directly rather than assuming either answer. Policies differ, and some are adding AI-related language. What’s consistent across carriers is that coverage responds to what you did, not what tool you used — which is another reason the review step matters more than the tool choice.
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.