ChatGPT for Real Estate Agents: Copy-Paste Prompts and a Fair Housing Setup (2026)

The short answer

ChatGPT is useful to a real estate agent in 2026 as a drafting tool: listing descriptions from your notes, replies to new leads, follow-ups to quiet ones, and recaps after showings and open houses. It goes wrong in two predictable ways, inventing facts about a property and describing people instead of the home, so the prompts below are built to catch both. Plus at $20 a month is enough for one agent; a brokerage with shared prompts or client documents belongs on ChatGPT Business.

This is the prompt kit. If you want the bigger picture first (what a team or brokerage should roll out, and in what order), that's in the AI for real estate hub. Everything here is copy-paste: swap the brackets for your details and go.

Set it up once before you prompt anything

Most bad ChatGPT listings come from a blank chat with no context. Put this in a project's instructions (or in your custom instructions on a personal plan) and every draft starts from it:

I'm a real estate agent in [CITY / MARKET]. I work mostly with [BUYERS / SELLERS / BOTH] in [PRICE RANGE OR AREA].

My voice: [3 adjectives], short sentences, no exclamation points, no "nestled", "boasts" or "dream home".
Here are two listings I wrote that sound like me: [PASTE TWO].

Rules for every draft:
- Only state facts I gave you. If something is missing, write [CHECK: ...] instead of guessing.
- Describe the property, never the people who should live there or the people who live nearby.
- Never mention schools as "good" or "bad", never describe a neighborhood's residents, never say who a home is "perfect for".
- Never quote a price, a rate or a payment.
- I approve everything before it's sent or posted.

That third rule is doing most of the fair-housing work, and the first one is doing most of the accuracy work. Both get their own section below.

Listing description prompts

1. The first draft, from your notes.

Write an MLS listing description from my notes below. Under [CHARACTER LIMIT] characters. Lead with the one thing a buyer would remember after a showing. Only use facts in my notes; mark anything you'd want to add but can't confirm as [CHECK: ...].

Notes: [beds/baths, square footage from the tax record or measurement, updates with years, lot, what conveys, what you noticed walking through]

2. The fact list, before you post. Run this on any draft, yours or ChatGPT's:

List every factual claim in this listing as a checklist: features, measurements, dates, what's included, location claims. For each, say whether it appears in my notes above. Flag anything that doesn't.

Listing: [PASTE]

This is the prompt that catches "recently renovated" when your notes said "new dishwasher." It takes 20 seconds and it's the one most agents skip.

3. The fair-housing second read.

Review this listing and flag any word or phrase that describes people rather than the property: who the home suits, who lives nearby, religion, race, national origin, disability, sex, families or children, age. Quote each phrase and suggest a version that describes the space instead. Don't rewrite anything else.

Listing: [PASTE]

Follow-up prompts

4. Reply to a new lead.

A lead just asked about [ADDRESS] through [ZILLOW / MY SITE / A SIGN CALL]. Their message: "[PASTE]". Draft a reply under 80 words that answers their actual question, offers two specific showing times ([TIMES]), and asks one question about what they're looking for. Sound like a person texting, not a brochure.

5. The quiet lead, two weeks later.

This buyer went quiet 14 days ago. What I know: [FIRST NAME], looking in [AREA] around [RANGE], last saw [ADDRESS] and said "[WHAT THEY SAID]". Draft a short check-in that gives them one new, useful thing (a new listing that fits, a price change, a question they asked that I can now answer) and makes it easy to say "not now". No "just checking in".

6. After a showing.

Turn my showing notes into a client update for [FIRST NAME]: what they liked, what they didn't, my honest read, and the next step. Under 150 words. Notes: [PASTE YOUR VOICE MEMO TRANSCRIPT OR NOTES]

Open-house recap prompts

7. The seller update.

Write my seller a recap of Sunday's open house at [ADDRESS]. Facts: [NUMBER] groups, [NUMBER] with agents, comments I heard: [LIST], questions people asked: [LIST]. Separate what we heard from what I recommend. Keep any price conversation for our call; don't suggest a number.

8. Attendee follow-ups, in one pass.

Here are open-house sign-ins with my notes: [FIRST NAMES + ONE LINE EACH, NO PHONE NUMBERS OR EMAILS]. Write a two-sentence follow-up for each that mentions the specific thing they asked about or noticed. Group them: already working with an agent / looking now / just curious. For anyone with an agent, write a note that respects that.

9. The neighbor note.

Write a short note to neighbors on [STREET] about the open house we just held: how many people came through and an offer to tell them what their home might be worth. Describe the street and the homes only, never the people who live there.

Notice prompt 8 leaves out phone numbers and emails. The draft doesn't need them, so they don't go in. That's the client-data rule in one line.

The fair housing rules of thumb

The law itself is short. The Fair Housing Act makes it unlawful "to make, print, or publish, or cause to be made, printed, or published any notice, statement, or advertisement, with respect to the sale or rental of a dwelling that indicates any preference, limitation, or discrimination based on race, color, religion, sex, handicap, familial status, or national origin" (42 U.S.C. 3604(c)). Many states and cities add protected classes of their own. "ChatGPT wrote it" doesn't change who published it.

The rules of thumb that follow from that:

• Describe the property, not the buyer. "Three bedrooms on one level" is a fact. "Perfect for a young family" or "ideal for retirees" is a statement about who should live there.
• Describe the place, not the neighbors. Distance to a park is fine. Words about who lives in the area are not your business in a listing.
• Watch the soft words. "Exclusive", "private community" and comments about churches or schools are where drafts drift. When the second-read prompt flags one, rewrite it around the feature.
• Your broker decides the hard calls. A chatbot isn't a compliance review, and neither is this post.

The client-data rules of thumb

• Never in a prompt: Social Security numbers, bank statements, pre-approval letters, tax returns, lockbox codes and showing instructions.
• Use first names. A follow-up draft doesn't need a full name, an address history or a phone number.
• Know where the conversation lives. On a personal plan, the history is yours and leaves with you. On a brokerage's Business workspace, it stays with the brokerage, and OpenAI says it doesn't train its models on Business data by default.
• Write it down. If you're a broker, these rules belong in a one-page policy every agent signs. My AI acceptable use policy template is a fine starting point; add the real estate lines above.

When should a brokerage move to ChatGPT Business?

Here are the plans OpenAI lists today, as of October 6, 2026:

Move to Business when any one of these is true: two or more people are sharing prompts or templates, agents are pasting anything from a transaction file, an agent has left and taken a year of client conversations with them, or you want to see who's using it. A 10-agent brokerage is $200 a month billed annually or $250 billed monthly. The full plan-by-plan detail is in my ChatGPT Business pricing breakdown.

Managing rentals too? Tenant messages are a different set of prompts; AI for property management covers that side.

Who can set up ChatGPT for a real estate team?

This is what I do. I'm Justin McKelvey, an AI consultant in Austin, Texas, and I set up ChatGPT Business for teams of 10 to 200, real estate teams and brokerages included. The ChatGPT Business setup starts at $4,500 and takes about two weeks: the workspace and seats, up to five apps connected, a shared project with your market and voice (the setup above, done properly), one workflow, a 60-minute training for the agents, a plain-language handoff doc and 30 days of support. You pay OpenAI for the seats directly. Not sure ChatGPT is the right tool? The $2,500 AI Readiness Assessment answers that first, and either way the first step is a free 30-minute call.

What to do today

Paste the setup block into a project, then run prompt 2 on your last three listings. If it finds a claim you can't back up, you've just found the reason to keep a human reading every draft.

Sources

• Fair Housing Act, 42 U.S.C. 3604, via Cornell Law School's Legal Information Institute, law.cornell.edu/uscode/text/42/3604 (read October 6, 2026)
• U.S. Department of Housing and Urban Development, "Housing Discrimination Under the Fair Housing Act", hud.gov (read October 6, 2026)
• OpenAI, ChatGPT pricing, chatgpt.com/pricing (read October 6, 2026)
• OpenAI, ChatGPT Business pricing, openai.com/business/pricing (read October 6, 2026)
• OpenAI, business data privacy page, openai.com/business-data (read October 6, 2026)