ChatGPT Wrote the Listing and the Agent Owns the Mistake
AI can draft the marketing but it cannot take responsibility for an inaccurate MLS entry. Advertising rules, virtual staging limits, fair housing and client confidentiality.
Artificial intelligence has become a routine real estate tool. Agents use it to draft listing descriptions, social posts, follow-up emails and market summaries, and the efficiency is real. So is the trap: an AI tool will state an invented fact with exactly the same confidence it uses for a verified one.
Here is the scenario I keep seeing. An agent feeds an old listing into a chatbot and asks for a polished luxury description. What comes back says the home has impact windows, a new roof, Gulf access and a fully permitted renovation. The prose is genuinely good. Two of those statements are not true.
The listing goes live. The buyer reads it, relies on it, and closes. And when the problem surfaces, neither the agent nor the brokerage can hand the responsibility back to the software.
The Advertising Rules Did Not Change
Florida law authorizes discipline against licensees for misrepresentation, concealment, false promises or culpable negligence, and for advertising property in a manner that is fraudulent, false, deceptive or misleading. Notably, a consumer does not have to prove a financial loss for that exposure to exist.
The National Association of Realtors (NAR) Code of Ethics runs parallel: Realtors must avoid exaggeration, misrepresentation or concealment of pertinent facts about the property, and must present a true picture in their advertising and marketing.
Neither of those standards contains an exception for words the licensee did not personally type. The duty attaches to publication, not authorship. Whether the description came from the agent, an unlicensed assistant, a marketing vendor or a language model, the name on the listing is the name that answers for it.
Separate Fact Gathering From Writing
The single most effective fix is structural. Stop asking the AI to describe the property, and start asking it to write from facts you have already verified.
The workflow looks like this:
Build a property fact sheet first, from reliable sources: the seller’s disclosure, the permit history, property appraiser records, the survey, prior closing documents, and invoices or warranties for any claimed improvement.
Resolve conflicts before drafting. If the disclosure says septic and the tax record says sewer, settle that question now — not after the listing publishes.
Instruct the tool to use only those facts and to leave gaps blank rather than filling them in. A model asked to write attractive copy will supply plausible details unless told not to.
Check every property-specific statement against the file after the draft comes back.
Seller-supplied information helps identify what needs verifying, but repeating it does not automatically insulate an agent from an inaccurate advertisement. “The seller told me” is a starting point for investigation, not a defense.
Words That Carry Legal Weight
Some marketing terms are not marketing at all. They are factual or legal claims wearing adjective clothing. Treat each of these as requiring documentary support before it appears in a listing:
“New” roof, “new” AC, “new” anything
“Permitted” or “fully permitted”
“Fully renovated”
“Impact” windows or doors
“Gulf access” or “direct access”
“Legal bedroom” and bedroom counts generally
“No flood damage” or “never flooded”
“Short-term rental permitted”
Square footage, lot size and year built
If the support is uncertain, investigate it or remove the claim. “Buyer to verify” is a defensible qualifier. An unqualified assertion is a representation.
Virtual Staging Has Quietly Become Something Else
Virtual staging used to mean adding a sofa to an empty room. Everyone understood the convention, and no one was deceived.
Current image tools do considerably more. They will remove a wall, enlarge a window, erase a stain, repair visible damage, brighten a canal into a water view, or add a pool to a yard that does not have one. The moment an edit changes a permanent feature or conceals a defect, the image has stopped depicting the property’s present condition.
Follow the MLS rules: label virtually staged and conceptual images clearly, and retain the original photograph where retention is required. But understand the limit of a label. Writing “virtually staged” under an image that hides a structural crack, or that manufactures a view the property does not have, can still leave a misleading overall impression — and the overall impression is the standard that matters.
Two Risks Agents Tend to Miss
Client confidentiality
Agents are now pasting contracts, inspection reports, financial statements and client emails into AI tools to summarize them. Tool terms vary widely, and some consumer products may retain prompts or use submitted content to improve the model.
A brokerage policy should name the approved tools and prohibit confidential, personal or financial client information from being entered into anything outside that list. And redacting a name is often not enough — an address, a parcel number and a closing date can identify a client and a transaction just as effectively.
Fair housing
AI-generated copy needs a fair-housing read every single time. Language about who a home is “perfect for,” or neighborhood descriptions that track protected characteristics, create exposure regardless of which tool produced the phrasing. Models reproduce the patterns in their training data, and some of those patterns are exactly the ones fair-housing law exists to prevent. Automation does not eliminate discriminatory effect, and it does not shift responsibility for publication.
A Human Publication Checklist
Before anything goes live:
Compare the text against the transaction file, line by line
Verify every property-specific claim against a source document
Read for fair-housing problems
Inspect every image that was altered, and confirm the labeling
Confirm no confidential information went into an unapproved system
Brokerages should document which tools are approved, what review is required, and who has authority to publish. That documentation is also what demonstrates reasonable supervision if a complaint is ever filed.
AI is genuinely useful for speed, structure and language. Factual verification, professional judgment and accountability stay with the people whose names appear on the listing.
Frequently Asked Questions
Is it against the rules to use AI to write listing descriptions?
No. Using AI as a drafting tool is not prohibited. What is prohibited is publishing false, deceptive or misleading advertising, and that standard applies identically no matter how the words were produced.
If the AI made up a fact, is the agent still responsible?
Yes. Responsibility attaches to the licensee who published the advertisement. A software error is not a recognized defense to a misrepresentation claim or a licensing complaint.
Does labeling an image “virtually staged” solve the problem?
It helps, and MLS rules generally require it, but it does not cure every alteration. A disclaimer does not make an image accurate if the edit removed a defect or created a feature that does not exist. The test is the impression the marketing creates as a whole.
Can I put a client’s contract into ChatGPT to summarize it?
Only if your brokerage has approved that tool and its terms are consistent with your confidentiality obligations. Assume anything entered into a consumer product may be retained, and remember that removing the name does not always de-identify the file.
A buyer relied on an inaccurate listing. What now?
That turns on what was stated, what was known, what the buyer could readily observe, and what the contract and disclosures said. These claims move on deadlines, so an agent, brokerage or buyer in that position should have the listing, the file and the contract reviewed promptly.
Talk to a Southwest Florida Real Estate Attorney
Prell Spearing Law Firm represents buyers, sellers, agents and brokerages in real estate transactions and disputes throughout Lee, Collier, Charlotte and Sarasota Counties — including claims arising from listing accuracy, disclosure and advertising.
If you are building an AI policy for your brokerage, or dealing with a listing that said something it should not have, contact the firm.
This article provides general information about Florida law as of the date of publication. It is not legal advice and is not a substitute for advice from counsel regarding a specific transaction. Case examples are composites with identifying details changed.