The short verdict
ChatGPT is a great writer and a bad researcher. Use it for words. Don't use it for numbers.
Use it for
- First drafts of listing copy from facts you supply
- Social captions, reels scripts, newsletters
- Follow-up emails and texts
- Objection-handling and role-play practice
- Turning your notes into a clean summary
Don't use it for
- Pulling comps or pricing a home
- Building a client-ready CMA
- Market stats you haven't verified
- Finding homes likely to list
- Anything with clients' private data
20 prompts that actually work
Each prompt gives ChatGPT a role, the facts, a format and limits. Replace anything in [brackets]. Hit copy, paste, edit.
Listing copy 4 prompts
Act as a listing copywriter for a licensed agent in [city, state]. Using ONLY the facts below, write a [150]-word MLS description. Lead with the single strongest feature. Describe the property, not the ideal buyer. Do not use words that reference protected classes (family, singles, adults, church, etc.). Do not add any feature, distance or amenity that isn't listed. At the end, list any fact you were unsure about. Facts: [beds, baths, sq ft, lot, year, upgrades, HOA, notable features]
Rewrite this listing description three ways: (1) MLS — factual, 120 words; (2) Instagram — warm, 60 words, 3 relevant hashtags; (3) luxury print — elegant, 90 words. Keep every fact identical across all three. Don't add features. [paste description]
Here are 8 features of a listing. For each, write one short line that turns the feature into a lifestyle benefit without describing who should live there. [list features]
Review the listing text below for language that could raise Fair Housing concerns — references to protected classes, the "ideal" buyer, religious landmarks, or neighborhood character. Quote each phrase, explain the concern, and suggest a neutral rewrite. If nothing is flagged, say so. [paste text]
Social & content 4 prompts
Write copy for a 6-slide Instagram carousel for a new listing. Slide 1: hook under 8 words. Slides 2–5: one feature each, under 15 words. Slide 6: call to action to DM me for a private showing. Facts: [facts]
Write a 30-second walkthrough reel script for [address/neighborhood]. Include on-screen text for each shot, a spoken line under 12 words per shot, and an opening line that stops the scroll. Tone: [confident / warm / playful].
I'm a real estate agent in [market] focused on [buyers/sellers/luxury/first-time]. Plan 12 social posts for the next 4 weeks: 4 educational, 4 local-lifestyle, 2 personal, 2 listing or sold. Give each a hook and a one-line caption idea. No market statistics — I'll add verified ones.
Turn these rough notes into a 250-word monthly email newsletter for my sphere. Friendly, not salesy. One local highlight, one tip, one soft call to action. Use only the numbers I give you. [notes]
Clients & follow-up 5 prompts
Write a text reply (under 300 characters) to a new online lead who asked about [property]. Thank them, answer [their question] only if I've given you the answer, and offer two specific showing times: [times].
Write a follow-up email to buyers I showed [number] homes today. Recap each home in one line using my notes, ask which they'd like to see again, and suggest a next step. Notes: [notes]
Write 5 different short check-in texts to past clients who bought 1–3 years ago. No asking for referrals directly. Each should feel personal and reference something they might care about as homeowners ([season/local event]).
Write a calm, clear weekly update email to my sellers. This week: [showings], [feedback themes], [online views]. Recommend [next step] and explain why in two sentences. Don't speculate about value.
Explain [escrow / contingencies / appraisal gaps / closing costs] to a first-time buyer in under 150 words, plain English, no jargon. End with: "Ask me anything — I'll walk you through your specific situation."
Scripts & negotiation 4 prompts
Role-play a homeowner who says "[objection — e.g. we'll just list with the agent who quoted 1% less]". Push back realistically three times. After the role-play, critique my answers and suggest stronger ones.
Write a 60-second opening for a listing appointment that builds trust, sets an agenda, and asks the sellers what matters most to them. Conversational, not scripted-sounding.
Help me prepare for a conversation recommending a price adjustment. Facts: [days on market, showings, feedback, comparable activity I've verified]. Give me an opening line, three talking points tied to those facts, and responses to two likely pushbacks.
Summarize these offers in a simple table for my sellers: price, financing, down payment, contingencies, closing date, escalation, notable terms. Then list three questions sellers should consider. Don't recommend one. [paste offer terms]
Your business 3 prompts
Write a 100-word agent bio in first person from these facts: [years, markets, specialties, a personal detail]. Confident, specific, no clichés like "passionate" or "dedicated".
Write a same-day follow-up email to open house visitors at [address]. Thank them, include one detail about the home, ask if they're working with an agent, and offer to send similar homes.
I have [hours] this week. Priorities: [list]. Build a day-by-day plan with time blocks for prospecting, client work, content and admin. Keep prospecting before noon.
Where ChatGPT breaks
These aren't prompt problems. No prompt fixes them, because they're about data ChatGPT doesn't have.
It can't see your MLS
No closed comps, no pending sales, no days-on-market, no off-market data. The Zillow app inside ChatGPT lets consumers browse listings, but it's a buyer search experience that sends people back to Zillow — not a connection to your MLS.
Workaround: paste comps you pulled yourself and ask only for the narrative.
It makes numbers up
Ask for a price per square foot or a "recent sale" and it may answer confidently with figures that don't exist. In a CMA, one invented comp is enough to lose a listing — or your credibility.
Workaround: tell it to use only facts you provide and to flag anything it's unsure of.
It drifts into Fair Housing trouble
Trained on decades of listings, it reaches for phrases like "perfect for families" or "walking distance to church." Your name is on the listing — not OpenAI's.
Workaround: prompts 1 and 4 above, plus your own review every time.
It doesn't watch your market
It won't alert you to a price drop, a new listing in your farm, or a tax reassessment, and it can't tell you which homeowners are likely to sell next.
Workaround: none inside ChatGPT. This needs a tool connected to live data.
It isn't built for client data
Pasting clients' names, finances and contract terms into a general chatbot is a privacy decision your broker should make, not one to make by default.
Workaround: strip identifying details, and check your plan's data settings and brokerage policy.
Keep ChatGPT for words. Get MLS-connected AI for the rest.
If breaks 1, 2 and 4 cost you time every week, you've outgrown a general chatbot. Sara AI is an assistant built only for agents and connected to live MLS and public-records data — it's what Douglas Elliman runs brokerage-wide as Elli, and individual agents can sign up at a public price.
Pricing: Lite $50/month · Pro $200/month · Enterprise custom.
Try Sara AI — 50% off your first monthApplies automatically through this link, or use code ASHER at checkout.
Other MLS-connected assistants exist — Sidekick (often through MLS or association partnerships) and Rechat's Lucy (usually bought by brokerages). If your brokerage already provides one, start there.
Frequently asked
Can ChatGPT do a CMA?
Not reliably. It isn't connected to your MLS, so it can't pull verified closed comps or adjust them. It can write the narrative around comps you paste in. MLS-connected assistants like Sara AI build the CMA itself in about 15 seconds.
Does ChatGPT have MLS access?
No. The Zillow app in ChatGPT lets consumers browse Zillow listings in a chat and routes them back to Zillow. It doesn't connect to your MLS account, pull closed comps or monitor your market.
Is ChatGPT safe for listing descriptions?
With review, yes — 75% of agents who use AI use it for listing copy, per NAR's 2026 Technology Report. Give it only verified facts, have it describe the property rather than the buyer, and check every claim before publishing.
ChatGPT free or Plus for agents?
Free is enough to test the prompts on this page. Agents who use it daily usually move to Plus ($20/month) for higher limits and better models. Neither tier adds MLS access.
What's better than ChatGPT for real estate?
For writing, ChatGPT is fine. For comps, CMAs, pricing, alerts and finding likely sellers, use an MLS-connected assistant. Sara AI is Lite $50/month or Pro $200/month; code ASHER takes 50% off the first month.
How many agents use ChatGPT?
NAR's 2026 Realtors Technology Report found 91% of agents who use AI use ChatGPT, up from 58% the year before, and 23% of agents use AI every day.
Sources
- NAR — The tech real estate agents rely on most in 2026
- Inman — NAR technology report on AI adoption (Sep 2026)
- Zillow Group — Zillow debuts the only real estate app in ChatGPT
- Florida Realtors — Fair housing and AI-written property descriptions
- The Close — ChatGPT pricing for agents
- Sara AI — features and pricing