ChatGPT Prompts for Real Estate Agents That Actually Work

By Emily Terrell — Top Coach and Speaker at Tom Ferry International. Speaker for NAHREP, eXp Con, and brokerages nationwide.

The best ChatGPT prompts for real estate agents give the model four things: your context, your real numbers, one job, and a review step. Generic prompt packs fail because they skip the first two. Below are the 10 prompts I run in my own business, each written to work in ChatGPT or Claude.

Key Takeaways

• The prompt matters less than the context behind it: your market, your voice, and your real numbers.
• ChatGPT and Claude both run a well-built prompt, so choosing a tool isn’t what’s holding your output back.
• Every prompt below works from a client profile, never a client’s name, contact details, or finances.
• The most valuable instruction you can add is asking the AI to separate facts you gave it from facts it supplied.
• Ten prompts tied to tasks you repeat every week will outproduce a downloaded pack of hundreds.

What is a real estate AI prompt?

A real estate AI prompt is a written instruction you give ChatGPT, Claude, or another AI assistant to complete one specific task in your business, like drafting a seller update or preparing for a listing appointment. A strong prompt includes your business context, the real data for the task, the format you want back, and the rules the output has to follow. A weak prompt is a one-line request that forces the model to guess. The guessing is where the generic copy comes from.

Why this matters for real estate agents

Most agents already have the tools. According to NAR’s 2025 Technology Survey (September 2025), ChatGPT is the most widely used AI tool among Realtors at 58%, followed by Gemini at 20% and Copilot at 15%. The same survey found only 17% of agents reported a significantly positive impact from AI, while 46% said it had no noticeable impact on their business.

Adoption has kept climbing since. According to RPR’s February 2026 survey of 225 NAR members, 82% of agents currently use AI and 34% save four or more hours a week. In that same survey, 63% named accuracy of outputs as their top concern.

Read those numbers together. Same tools, wildly different results. The agents saving four hours a week aren’t using a secret app. They’re giving the model better instructions and better inputs, and they’ve built a review step for the accuracy problem instead of hoping it doesn’t show up. NAR Deputy Chief Economist Jessica Lautz put the priority plainly in the survey release: what matters most is still “the trusted relationship between the agent and client.” Your prompts should protect that relationship, not outsource it.

Is ChatGPT or Claude better for real estate prompts?

Neither, for the prompts in this post. Here’s the thing nobody wants to tell you: the tool debate is a distraction from the documentation problem underneath it.

Both companies say the same thing in their own guidance. OpenAI’s prompting best practices tell you to be specific and detailed about context, outcome, length, format, and style. Anthropic’s prompting guide for Claude tells you to treat the model like a brilliant new employee who doesn’t know your norms or workflows yet. Neither tells you to hunt for a magic phrase.

So the real question isn’t which tool. It’s whether you’ve written down what a new employee would need to know about your business. That document is the context block.

What goes in the context block that powers every prompt?

The context block is a short standing brief you paste once at the top of a conversation, or save permanently in a Claude Project or a Custom GPT. Every prompt below assumes it’s already loaded.

CONTEXT

I’m a licensed real estate agent with [Brokerage Name] serving [city and submarkets].

I mostly work with [property types and price bands].

My voice: short sentences, contractions, direct, no exclamation points.

Here are three things I actually wrote: [paste 3 samples].

Never use: dream home, nestled, boasts, rare find, must-see, [your list].

Fair housing: describe properties, never people. Never say who a home is

“perfect for.” Never use schools or places of worship as lifestyle selling

points. Never use “safe” or “family-friendly.”

Facts: use only facts and numbers I provide. If something is missing,

write [VERIFY] instead of guessing.

Client data: I will never give you names, phone numbers, emails, addresses

tied to a person, or financial details. Work from the profile I provide.

If you haven’t written your voice samples and banned list yet, start with the full build in how to train AI on your brand voice. It takes about twenty minutes and it’s the highest-return step in this entire post.

The 10 ChatGPT prompts for real estate agents

Each prompt below is built on a profile, not a person. For the content prompts I didn’t repeat here, like listing descriptions and Instagram captions, see how to use ChatGPT for real estate content creation.

Prompt 1: How do you re-engage a lead who went quiet?

Give the model the last real thing the lead told you, and ask for three angles instead of one message.

Using my context, write 3 text messages to re-engage a lead who went quiet.

Profile: [buyer or seller], [price band], [submarket], last contact

[number] days ago. Last thing they told me: [one line, no names].

Rules: under 300 characters each. One question per text. No “just checking

in.” No urgency I can’t back up with a fact.

Give me 3 options with different angles: a market fact, a useful resource,

and a direct question.

Three angles let you match the message to the person, which a single draft can’t do. Send only to contacts you have permission to text.

Prompt 2: How do you prep for a listing appointment with AI?

Ask for questions and likely objections, not a script.

I have a listing appointment on [date].

Seller profile: [property type], [submarket], [timeline], reason for

selling in 3 words: [reason]. What they’ve said about price: [summary].

Give me:

1. The 5 questions most likely to surface their real motivation and timeline.

2. The 3 objections this seller is most likely to raise, based only on

   this profile.

3. What I should bring to answer each one.

Don’t write a script. Give me talking points I’ll say in my own words.

The output is a prep sheet. You walk in with better questions instead of a speech.

Prompt 3: How do you write a weekly seller update in five minutes?

Feed it the week’s activity and force a structure that ends in a decision.

Turn this week’s activity into a seller update email.

Data: [showings], [online views if available], [feedback themes,

paraphrased], [new competing listings], [nearby price changes].

Structure: what happened, what it means, one recommended next step,

one question for the seller.

Under 180 words. Plain English. Use only the numbers above. If the data

points toward a price conversation, say so directly and calmly.

Sellers don’t fire agents over bad weeks. They fire agents over silent weeks.

Prompt 4: How do you prepare for a price adjustment conversation?

The AI organizes the evidence. You do the talking.

Help me prepare for a price conversation. I’ll do the talking; you

organize the evidence.

Data: list price [ ], days on market [ ], showings [ ], feedback

themes [ ], active competition [price, sq ft, days on market for each,

no addresses], recent solds [same format].

Give me:

– The 3 strongest data points, ranked, each in one sentence a seller

  would understand.

– The question I should ask before I mention a number.

– Two options to present, with the trade-off of each.

Use only my numbers. Flag anything I should verify in the MLS before

the call.

Asking for the question before the number is the part most agents skip. It turns a verdict into a conversation.

Prompt 5: How do you rehearse objections with AI?

Make the model play a skeptical client, then make it grade you.

Role-play a [seller or buyer] with this profile: [profile].

Raise this objection: “[objection].”

Push back the way a skeptical real person would. Don’t make it easy.

After 4 exchanges, stop and grade my responses: where I was clear,

where I got defensive, and one line I should replace.

This is the closest thing to a role-play partner at 9pm. Run it the night before any appointment you’re nervous about.

Prompt 6: How do you turn market stats into an email your sphere will read?

Pick one number that matters and explain it like a neighbor.

Turn these numbers into a 120-word market email for my sphere in

[submarket].

Data: median price [ ] and change vs. last year [ ], active inventory [ ],

median days on market [ ], months of supply [ ].

Source: [MLS or report name, month].

Lead with the one number that matters most to a homeowner this month and

why. One call to action: reply for a home value review.

Don’t predict prices. Don’t add numbers I didn’t give you.

“Don’t predict prices” isn’t caution for its own sake. A forecast your AI invented is a forecast with your name on it.

Prompt 7: How do you summarize showing feedback for a seller?

Strip the names, find the pattern, sort it by what the seller can actually do.

Here’s this week’s raw showing feedback, pasted from agent emails and

texts: “””[paste]”””

Remove agent names and anything identifying. Group the feedback into

themes. For each theme: how many showings mentioned it, and whether it’s

something the seller can fix, price for, or ignore.

Keep it neutral. No softening, no alarm.

One comment is an opinion. Four comments on the same thing is a decision waiting to be made.

Prompt 8: How do you write a past client note that doesn’t feel automated?

One true detail does more than any template.

Write a short note to a past client on their [1, 2, or 5]-year home

anniversary.

Profile: bought a [property type] in [submarket] in [year].

One true detail from their purchase: [detail, no names or financials].

Under 80 words. Warm, specific, zero sales pitch. End with a question

they’d actually want to answer.

Give me 2 versions: one for a handwritten card, one for a text.

You don’t need more leads. You need a better system for the people who already trust you.

Prompt 9: How do you write a 30-second video script with AI?

Write for the ear, not the eye.

Write a 30-second video script I’ll record on my phone about

[listing, market stat, or process topic].

Facts: [paste].

Format: hook in the first sentence (a number, a question, or a bold

statement), 3 short points, one call to action to DM me.

Write it to be said out loud: short sentences, no words I wouldn’t say

in conversation. Mark where I should pause.

If you trip over a line on the first read, cut it. The script should sound like you on your best day, not like a brochure.

Prompt 10: How do you check an AI draft before you publish it?

Make the AI audit its own work before you do.

Review the draft below before I publish it.

1. List every factual claim and mark whether it came from information

   I provided it or from you.

2. Flag any word or phrase that describes people instead of the property,

   or that could signal a preference for a type of buyer.

3. Flag anything that doesn’t match my voice samples.

4. Confirm my broker name is present: [Brokerage Name].

Don’t rewrite yet. Give me the list first.

Draft: “””[paste]”””

This one runs after every other prompt on this list. Federal fair housing rules make it unlawful to publish a housing ad that indicates a preference or limitation based on a protected class, and that includes the words and phrases you choose. The rule applies to what you publish, not to who drafted it. For the full 90-second review sequence, including broker identification, use the AI listing description compliance checklist.

This is general information, not legal advice. Fair housing and advertising rules vary by state, MLS, and brokerage. Confirm your obligations with your broker or a licensed attorney.

How I use this in my own business

The prompt I run most is number three, the weekly seller update. Every listing in my business gets one. Before I built the prompt, those updates were the first thing to slip in a heavy week.

Now my assistant pulls the week’s showing count, feedback, and new competition into a profile with no names attached and runs it inside my Claude Project, where the context block already lives. I get the draft, fix anything that needs a licensee’s judgment, and send it myself. On a Stone Oak listing this summer, the update surfaced the same comment across four showings: the secondary bedrooms read small in the photos. That one paragraph turned what would have been a price conversation into a photography reshoot.

The seller didn’t need a pep talk. They needed the pattern, in writing, every week. That’s the system working.

“A prompt pack gives you someone else’s words. A context block gives the model yours. Ten prompts built on your own market, voice, and numbers will outproduce a thousand borrowed ones, and they’ll keep getting better every week you use them.”
— Emily Terrell, Tom Ferry Coach

Common mistakes

Shopping for prompts instead of writing context. The download feels like progress. The context block is the progress.

Pasting the client file. Names, loan amounts, and reasons for moving never go into a chat window. The model does the same work from a profile. The full rule set is in is it safe to put client data in ChatGPT.

Asking for numbers the model doesn’t have. If you didn’t paste days on the market, the AI will invent a plausible one. Every prompt above says “use only my numbers” for that reason.

Publishing the first draft. Prompt 10 exists because fluent copy is the easiest kind to trust and the hardest kind to catch.

Rewriting prompts for every tool. If a prompt only works in one app, it’s leaning on the tool instead of your context. Fix the context and the prompt travels.

Frequently Asked Questions

Do ChatGPT prompts work the same in Claude?

Yes. A well-built prompt is plain-language instructions plus your context, and both tools follow that structure. The output style differs slightly, so test the same prompt in each and keep the version that sounds more like you. What changes the result far more than the tool is whether your context block, voice samples, and real numbers are in the prompt.

Should real estate agents buy a prompt pack?

Usually not. A pack of 100 generic prompts gives you someone else’s wording with none of your market, voice, or numbers behind it, which is why the output sounds like every other agent. Start with ten prompts tied to tasks you repeat weekly, add your context block, and refine each one after real use. That library will outperform any download.

Can I put client information in an AI prompt?

Not identifying information. Leave out names, phone numbers, email addresses, street addresses tied to a person, loan amounts, and reasons for moving. Replace them with a profile: property type, price band, submarket, timeline, and situation in a few words. The model doesn’t need to know who the client is to do the work, and your confidentiality obligation doesn’t pause for convenience.

How many AI prompts does a real estate agent actually need?

Fewer than you think. In my coaching, most agents repeat the same eight to twelve writing and prep tasks every week: lead follow-up, seller updates, market emails, appointment prep, and review. Build one strong prompt for each, save them where you’ll actually use them, and stop. A short library you run every week produces more than a long one you scroll past.

Can AI write a price reduction conversation for me?

It can organize the evidence, but it shouldn’t write the script you read to a seller. Give it your days on market, showing count, feedback themes, and comparable data, and ask for ranked talking points and options with trade-offs. Then say it in your own words. Sellers trust the agent who explains the numbers, not the one reading them.

Where should I save my real estate prompts?

Save your context block in a Claude Project or a Custom GPT so every conversation starts already briefed, and keep your task prompts in one shared document your assistant can access. Name each prompt by the job it does, not the tool. When you fix the same thing in a draft twice, add that fix to the saved prompt so the library improves every week.

Bring this to your team or event

Emily Terrell speaks at brokerage events, real estate conferences, and team trainings on AI, systems, and social media — the exact playbook in this post, delivered live to your audience. As a Top Coach and Speaker at Tom Ferry International and an active agent closing 70+ transactions a year, Emily speaks from the stage about what’s working right now, not theory. Recent stages include NAHREP and eXp Con.

Book Emily to speak at your next event:
Email: eterrell@yourcoach.com
Phone: (210) 400-9191
Web: coachemilyterrell.com

For real estate agents who want to implement this: Get the weekly real estate prompt library at weeklyrealestateprompts.com or follow @coachemilyterrell on Instagram for daily systems and AI breakdowns.