ChatGPT for Real Estate Listing Descriptions: Prompts, Fact Sheets and Fair-Housing Checks

Listing descriptions are short, formulaic and written under time pressure, which is why so many agents now draft them with ChatGPT or a similar tool. The quality of the result depends almost entirely on the facts the agent supplies. The legal risk depends on a handful of words the agent has to catch before the listing goes live.

This guide covers both: a fact sheet that gives the tool what it needs, three prompts, and a review pass built around the Fair Housing Act and truthful-advertising duties. Sources were checked on November 9, 2026.

Start with a fact sheet, not a prompt

When ChatGPT is asked to “write a listing for a 3-bed house,” it fills the gaps with stock phrases and invented details. A structured fact sheet stops that. Agents can keep this as a template and complete it at the listing appointment.

PROPERTY FACT SHEET
Type: [single-family / condo / townhouse]
Beds / baths: [X / X]
Interior square feet and source: [X, from appraisal / tax record / measurement]
Lot size: [X]
Year built: [X]
Recent updates with year: [e.g. roof 2023, kitchen 2024]
Standout features: [3 to 6 items]
Layout notes: [e.g. primary suite on main level, open kitchen and living area]
Accessibility features, stated factually: [e.g. step-free entry, elevator, wide doorways, or "none"]
Outdoor space: [X]
Parking: [X]
HOA and fees: [X]
Nearby, by distance or time: [e.g. 0.4 miles to a park, 10 minutes to downtown by car]
Things NOT to mention: [e.g. tenant-occupied, pending repairs disclosed separately]
MLS remarks limit: [X characters]

The “source” line on square footage matters. If the copy states a figure, the agent should know where it came from.

Three prompts

Prompt 1: the main description

You are drafting an MLS listing description for a licensed real estate agent.
Use only the facts in the fact sheet below. Do not add features, amenities or neighborhood claims that aren't listed.

[PASTE FACT SHEET]

Write a description of no more than [X] characters.
Lead with the strongest feature. Use specific nouns instead of adjectives where possible.
Describe the property and its location only. Do not describe who the home would suit or who lives nearby.
After the description, list any fact you were unsure how to use.

Prompt 2: versions for other channels

Using the approved description below, write:
1. A 280-character version for listing portals' short summary fields.
2. A 120-word version for the agent's website.
3. Two social captions under 50 words each.
Keep every factual claim identical to the approved version. Add nothing new.

[PASTE APPROVED DESCRIPTION]

Generating variations from an approved description, rather than from scratch each time, keeps the facts consistent across channels.

Prompt 3: a first-pass wording check

Review this listing description. Flag any word or phrase that:
- describes a type of buyer or occupant (for example by family status, age, religion, sex, disability, race or national origin);
- could suggest some people are unwelcome;
- makes a claim that would need a source (school quality, crime, investment returns, square footage).
Quote each flagged phrase and suggest a property-focused rewrite. Do not rewrite the rest.

[PASTE DESCRIPTION]

Prompt 3 is a helper, not a compliance review. It can miss things and flag harmless words. The agent, and the brokerage’s own review process where one exists, makes the call.

Fair housing: what the rules say

Section 804(c) of the Fair Housing Act prohibits housing ads that indicate a preference, limitation or discrimination based on a protected characteristic. HUD’s regulation at 24 CFR 100.75 covers race, color, religion, sex, handicap (disability), familial status and national origin. It applies to written and oral notices and statements, including the “words, phrases, photographs, illustrations, symbols or forms” used. HUD’s Fair Housing Act overview lists the same protected classes.

Two recent changes affect where agents find guidance, though neither changes the law:

  • HUD withdrew its 2024 digital-advertising guidance. A September 17, 2025 HUD memo withdrew several fair housing guidance documents, including the April 29, 2024 guidance on advertising through digital platforms, according to LeadingAge’s summary of the memo.
  • The 1995 advertising memo is harder to find. Oregon REALTORS reported in April 2026 that HUD removed its 1995 memo on advertising under Section 804(c) from its guidance pages, though the document can still be found online.

The statute and the regulation still apply in full. State and local fair housing laws can add more protected classes, so agents should check the rules where the property sits.

Rewrites that keep the copy about the property

These examples apply the regulation’s principle. They aren’t an official HUD list.

Draft phraseProblemProperty-focused rewrite
“Perfect for a young family”Signals a preference by familial status and age“Three bedrooms on one level, fenced backyard”
“Ideal for empty nesters”Signals a preference by familial status“Low-maintenance lot, primary suite on main floor”
“Not suitable for wheelchairs”Discourages buyers with disabilities“Three steps up to the front entry”
“Great bachelor pad”Signals a preference by sex and familial status“Open-plan one-bedroom with home office nook”
“Adults only”Familial status limitation, unless the property qualifies as housing for older personsRemove, or state the qualifying age restriction accurately

Stating accessibility features factually is useful information, not a preference. “Step-free entry and a roll-in shower” tells every buyer something true.

A worked example

Here is how a review pass changes a typical first draft. The property is an invented example.

First draft from a thin prompt:

Welcome home to this stunning chef’s kitchen and spacious family retreat, perfect for a growing family! Nestled in a safe, friendly neighborhood with top-rated schools, it’s an entertainer’s dream.

Problems: “perfect for a growing family” signals a familial-status preference. “Safe” and “top-rated schools” are claims that need a source, and the fact sheet gave none. “Chef’s kitchen” isn’t in the fact sheet either.

Revised from the fact sheet:

Three-bedroom, two-bath home with a 2024 kitchen update: quartz counters, gas range and a walk-in pantry. The main living area opens to a fenced backyard with a covered patio. Park 0.4 miles away; downtown about 10 minutes by car.

Every claim in the revision can be checked against the fact sheet, and none of it describes the buyer.

Truthful advertising

Fair housing isn’t the only check. NAR’s Code of Ethics (effective January 1, 2026), Article 12, requires REALTORS® to present a true picture in their advertising. AI tools are prone to small embellishments: “chef’s kitchen” for a standard one, “steps to the beach” for a 15-minute walk, “newly renovated” for work done years ago. Each claim in the final copy should trace back to the fact sheet. AI-edited or virtually staged photos raise the same question, and MLS rules on labeling them vary by market.

Don’t paste this in

Seller names, their reasons for selling or their timeline pressure, tenant names, alarm and lockbox codes, when the home will be empty, or any disclosure details the seller hasn’t approved for marketing. Copy shouldn’t say who a home suits by race, color, religion, sex, disability, familial status or national origin. Under 24 CFR 100.75, ads that indicate a preference or limitation on those grounds are unlawful, whoever or whatever wrote the first draft.

Quick answers

Does ChatGPT keep what an agent pastes in? On personal Free, Plus and Pro accounts, OpenAI’s data controls page says conversations may be used for model training by default unless the “Improve the model for everyone” setting is turned off. Business workspaces aren’t used for training by default.

Is the agent responsible for AI-written copy? Yes. Fair housing and advertising rules apply to the published ad, regardless of how it was drafted.

Do listings need to say AI was used? We couldn’t confirm a general national rule requiring it. Brokerage policies and MLS rules may set their own requirements.

Can the tool check the copy for compliance? It can flag candidate phrases, as in Prompt 3. It can’t replace the agent’s review or the brokerage’s.

For the rest of the deal cycle, from listing appointment to post-closing follow-up, see AI for real estate agents.



Previous Post
AI for Real Estate Agents: A Practical Playbook by Deal Stage