Ninety-one percent of agents who use AI are using ChatGPT, according to the National Association of Realtors 2026 Technology Report, and the top three uses are listing descriptions, social posts and follow-up email. These chatgpt prompts for real estate agents cover those three plus the ones that are more valuable and less used. Each includes the constraint that keeps the output usable, which is the part most prompt lists leave out.
One thing before the list. Every prompt below ends with a fair housing instruction, and that is not decoration.
Models produce phrases like "perfect for young families" and "great schools, quiet area" constantly, because those phrases are everywhere in listing copy. Familial status is a protected class and describing an area's character can constitute steering. The liability is yours. Read every description before it publishes.
ChatGPT prompts for real estate agents: listing copy
1. The description that does not steer
Write a 120-word listing description for the property details below. Describe only the property and its features. Do not describe who it would suit, do not mention schools, demographics, religious institutions or the character of the neighborhood, and do not use the words family, families, perfect for, or ideal for. Plain, concrete, no adjectives doing work a fact could do. Details: [paste]
The prohibition has to name the words. "Be fair housing compliant" is too abstract and the model will produce a violation anyway.
2. The rewrite that shortens
Cut this listing description to 90 words without losing any factual detail. Remove every adjective that is not load-bearing. Keep the same order of information. [paste]
Shortening is the thing models do best and agents do least.
3. The feature list from a walkthrough
Here are my rough notes from a walkthrough. Turn them into a clean bulleted feature list, grouped by floor. Flag anything that sounds like a defect so I can decide whether to include it. [paste]
The flag instruction is what makes this one useful. It surfaces the things you half-noticed.
Follow-up and communication
4. The follow-up that is not a check-in
Write a four-sentence follow-up email to a seller I met eight days ago. The purpose is to give them one specific useful thing, not to ask if they have thought about it. Context: [what you discussed]. The useful thing: [a comp, a timeline fact, a market note]. No pleasantries at the start. One clear next step at the end.
"Just checking in" is the most-sent and least-answered email in this business. Forcing the prompt to carry something useful fixes it.
5. The objection response
A seller said: "[objection]". Write three different responses. Each one should acknowledge the real concern underneath it, give a specific fact or mechanism, then ask a question that moves the conversation. Different angles, not three rewrites of the same answer. No pressure tactics.
The three-angles instruction is what stops you getting one answer phrased three ways.
6. The update nobody has to chase
Write a short status update for a client whose transaction is at [stage]. Say what has happened, what happens next, what I need from them, and when they will hear from me again. Four short paragraphs. No apology, no filler.
Saves the most time
Follow-up and status updates, because they are frequent and formulaic
Earns the most
Objection handling, because the quality varies most by rep
Needs the most review
Anything describing a property or an area
Preparation and analysis
7. The CMA narrative
Here are six comparable sales. Write two paragraphs explaining to a seller how these support a list price of [X]. Name each comp's relevant difference from the subject property. Do not overstate certainty. If any comp looks unusual to you, say so rather than smoothing it over.
The last sentence is the one that earns its place. It occasionally catches a comp you should not have included.
8. The pre-appointment brief
I have a listing appointment with [seller situation]. Give me the five questions I should ask, the two objections most likely to come up, and the one thing I should find out before I quote anything. Keep it under 200 words.
9. The document explainer
Explain what this document does, in plain language, for someone who has never bought a house. Six sentences maximum. Do not give legal advice or tell them whether to sign. [paste]
The prompt not to use
Asking a model to estimate a property's value, or to produce a comp set from its own knowledge.
It will give you a number. The number will be confident, formatted and wrong, because the model is not connected to current sales data and is pattern-matching from training text that may be years old. An agent who quotes it to a seller has made a representation they cannot support.
Use the MLS for data. Use the model for words about the data.
Good use of a model
Writing about comps you pulled
Explaining a document
Drafting a follow-up you send
Shortening your own copy
Bad use of a model
Producing comps from memory
Advising whether to sign it
Sending on your behalf without consent
Inventing a market statistic
Two habits that make all of these better
Paste real material. Prompts that work carry your notes, your comps, your actual objection. A prompt with no input produces generic output, and generic output is what makes AI-written copy recognizable.
Tell it what not to do. Models respond better to prohibitions than to aspirations. "No adjectives that a fact could replace" works. "Write compellingly" does not.
Frequently asked questions
What is the best ChatGPT prompt for a listing description?
One that names the forbidden words explicitly rather than asking for fair housing compliance in the abstract. Prompt 1 above lists family, families, perfect for and ideal for, because the model will produce them otherwise.
Can ChatGPT write a CMA?
It can write the narrative explaining comps you pulled. It cannot produce the comps. A model generating comparable sales from its own knowledge is inventing them.
Are these chatgpt prompts for real estate agents safe to use on client-facing copy?
With review, yes. The fair housing risk is real and the liability sits with you, so read every property or area description before it publishes.
Does it matter which AI tool I use?
Not much for writing. 91 percent of AI-using agents are on ChatGPT per NAR's 2026 report, and the prompts above work anywhere. What matters more is the second step, connecting something to your lead flow, which is a different kind of tool.
Why does AI-written copy sound like AI-written copy?
Usually because the prompt carried no real input. A model given your actual walkthrough notes writes differently from one given a property type.
Should AI send my follow-ups automatically?
Only with consent. AI drafting a message you send keeps you as the sender and stays outside the automated-messaging rules. AI sending on your behalf does not. See our post on
AI cold calling laws.
Where this goes next
These save time. The thing that changes your pipeline is what happens to a lead at 9pm on a Saturday, which is covered in AI for real estate agents.
If you want help connecting the writing to the lead flow, book a free strategy call. We go through your response times, your follow-up and your local visibility. It is a working session, not a pitch.
*Sources: National Association of Realtors 2026 Technology Report, published September 22, 2026. Fair Housing Act, 42 U.S.C. 3601 et seq. Last reviewed October 2026.*
