Almost nobody asks this out loud, which is exactly why it is worth answering properly. An agent drafts listing remarks with ChatGPT, edits them, publishes them, and then a small thought arrives: was I supposed to mention that?
The thought is worth taking seriously. But the question — do you have to tell your seller you used AI? — is two questions wearing one coat, and they have different answers.
Two questions, not one
The first is regulatory: am I legally or ethically required to disclose this? That has a reasonably clear answer and it depends entirely on what the AI touched.
The second is commercial: what happens to the relationship if my client finds out and I never mentioned it? That one has no rule attached, and it is the one that actually costs people business.
What genuinely requires disclosure
The obligations attach to what the buyer sees, not to how you produced it.
Altered images: yes, and increasingly by statute. Virtual staging, removed power lines, replaced skies, AI video generated from stills. Every major MLS requires disclosure of virtual staging in some form, and since 1 January 2026 California’s AB 723 requires a conspicuous label plus a link to the original unaltered image. The rules piece covers what each board asks for.
Misleading claims: yes, always, and this predates AI by decades. If the copy says something untrue about the property, it does not matter whether a person or a program wrote it. Article 2 of the NAR Code of Ethics puts that squarely on you.
Listing copy you wrote with help and then edited: generally no. Text you have reviewed, corrected and approved is your text. No rule requires you to itemise your tools, any more than you disclose the template you started from, the spell checker, or the colleague who read it back to you.
The line is consistent once you see it. Disclosure exists to stop a buyer being deceived about the property. It does not exist to audit your writing process.
Your brokerage may have its own answer
This is where agents get caught, because a brokerage policy is not the law but it is absolutely enforceable against you.
Many brokerages have now written AI policies, and they vary more than you would expect. Some require a disclosure line in the listing agreement. Some prohibit putting any client information into a public tool. Some restrict which tools may be used at all. Some have nothing written and would very much like you to ask before you find out the hard way.
One email to your broker, before you build a workflow on any of this. It costs a minute and it is the cheapest insurance in the article.
The question that actually matters
Set the rules aside for a moment. Imagine your seller learns, six weeks in, that the listing description was drafted by AI. Two versions of that moment exist.
In the first, they never heard you mention it. Now they are wondering what else was automated, whether you actually looked at their house, and what they are paying a commission for. Nothing improper happened, and the damage is real anyway.
In the second, you told them at the listing appointment, in one sentence, framed as competence rather than confession. They forgot about it immediately, because you had already made it a non-event.
The difference between those outcomes costs nothing and takes ten seconds.
The sentence to use
Somewhere in the listing presentation, where you talk about marketing:
“I use AI tools to speed up the first draft of the marketing copy, which frees up time for the parts that need me. Everything gets rewritten and checked by me before it goes anywhere, and every fact gets verified against the property. If we use any digital staging on the photos, that gets labelled, because the rules require it and because buyers should know.”
Three things are happening in that paragraph. You are telling them the truth, you are positioning it as efficiency rather than shortcut, and you are pre-empting the staging conversation before a buyer’s agent raises it. Sellers almost never have a follow-up question. The ones who do are usually curious rather than concerned.
What not to do: hide it and hope. In a year when every one of your sellers is using these tools at work themselves, being cagey about it reads worse than the thing you were being cagey about.
The issue nobody talks about
While everyone debates whether to mention the listing description, a genuinely serious problem goes unexamined: what you paste into the chat box.
Client names. Property addresses before they are public. Financial details from a pre-approval letter. The contents of an offer. Notes about a seller’s divorce, job loss or timeline pressure.
Your client never agreed to that, and unlike the question of who drafted the remarks, this one involves information they gave you in confidence going somewhere they cannot see. Consumer AI tools are not a confidential channel, whatever the settings page implies, and treating them as one is the real exposure in most agents’ AI workflow.
The habit that fixes it costs nothing: describe the situation without identifying the people. “A seller under time pressure who needs to close in 30 days” gets you exactly the same quality of draft as the version with their name and address in it.
A short summary
- Edited listing copy — no disclosure required in most places. Mention it anyway, once, at the listing appointment.
- Altered or staged images — disclosure required, by your MLS and in some states by statute. Not optional.
- AI video from stills — disclose. The buyer believes they are watching real footage.
- Anything misleading — your responsibility regardless of what produced it.
- Client data in public tools — the actual risk, and the one nobody is discussing. Strip identifying details every time.
One place where the answer is not a judgement call at all is the telephone. Using an AI voice to call people who never agreed to be called runs into federal rules rather than etiquette, and that is set out in the piece on AI cold calling.
Common questions
Does the buyer need to be told, or just my seller?
Image disclosure is aimed at the buyer and travels with the listing, which means it reaches everyone. The conversation about your drafting process is a relationship matter with your client, not a public notice.
What if a seller asks me not to use AI at all?
Then do not, and say so plainly. It is their listing and their preference is reasonable even if you disagree with it. Worth asking what specifically concerns them: usually it is a worry about generic copy or invented details, both of which you can address directly rather than by abandoning the tool.
Should I put an AI clause in my listing agreement?
Ask your broker; some brokerages already require one. A short, plain-language clause is defensible and takes the question off the table permanently. Do not draft it yourself out of an article on the internet, including this one.
Is it different for buyer clients?
The same principles apply, with one addition worth watching. If you use AI to help evaluate properties, summarise inspection reports or draft offer strategy, the confidentiality point becomes sharper, because buyer clients tell you things that would weaken their position if they travelled. Same fix: describe the situation, never the person.
The same reasoning applies to recording. If you run an AI notetaker on listing appointments, the consent rules differ by state and in about a dozen of them everyone on the call has to agree first.
General information, not legal advice. Disclosure rules vary by state, by MLS and by brokerage, and they are changing quickly. Check with your broker and, where the stakes justify it, an attorney licensed in your state.