ChatGPT for Real Estate Investors: 10 Prompts

ChatGPT for real estate is most useful on work where you supply the facts: turning call notes into a follow-up, building an underwriting sheet from numbers you pulled, prepping for a seller conversation, or writing the buyer blast for a house you have under contract. It's unreliable when it has to find or invent the facts itself, like comps, market rents or what a city ordinance says. Below is which plan to buy (prices checked on OpenAI's site in October 2026), ten prompts I'd hand any investor, and the habits that keep the output honest.

Which ChatGPT plan to buy

Source: ChatGPT pricing and the ChatGPT Business overview.

Plus is the one to buy if it's just you. Projects let you keep your buy box, your offer formula and your scripts in one place so you aren't pasting them into every chat. If a VA or acquisitions rep uses it with you, Business is worth the extra $5 a seat: OpenAI says it won't train on Business workspace data, while the individual plans list training as opt-out, so you'd have to remember to switch it off. Seller names, addresses and mortgage balances are exactly the data to keep out of model training.

Claude Pro is the same $20 a month ($17 billed annually), and every prompt below works in it too.

Set up a project with your buy box

Before the prompts, give ChatGPT the context it would otherwise guess at. Create a project and paste something like this into its instructions, with your own numbers:

I'm a real estate investor buying single-family houses in [MARKET].
Buy box: [PRICE RANGE], [BEDS/BATHS], built after [YEAR], no [EXCLUSIONS].
Exit strategies: wholesale assignment, or flip if the spread is over [AMOUNT].
My max offer formula: ARV x [PERCENT] minus repairs minus [ASSIGNMENT FEE].
Write in plain, short sentences. Never invent a number. If a figure I
need is missing, ask me for it instead of estimating it.

The last two lines do most of the work. Telling the model to ask for missing numbers is the best defense against it filling gaps with plausible ones.

ChatGPT prompts for real estate investors

These are grouped by where they fall in a deal. Replace everything in brackets.

1. Build an underwriting sheet from your numbers

Build a flip analysis table from these inputs. Show each formula.
Purchase: [PRICE]. ARV: [ARV] (from my comps, don't change it).
Repairs: [LIST WITH COSTS]. Holding months: [N]. Loan: [TERMS].
Closing costs buy and sell: [PERCENTS]. Then tell me the max offer
that still leaves a [AMOUNT] profit, and show the math.

Paste the table into a spreadsheet and check the formulas there before you trust the profit number.

2. Find the assumption that kills a rental deal

Here is my rental analysis: [PASTE NUMBERS]. Which single assumption,
if it moved by a realistic amount, would push monthly cash flow
negative first? Rank the top three and say how far each would have
to move.

3. Prep for a seller call

Here are my CRM notes on this seller: [PASTE NOTES]. Summarize what
we know about their motivation, timeline, condition and price in four
lines. Then list the five questions I still need answered, in the
order I should ask them, worded the way I'd say them on the phone.

4. Turn a call transcript into CRM notes

Here is a call transcript: [PASTE]. Extract: property address, owner
name, motivation, timeline, condition issues, mortgage or liens
mentioned, asking price, objections raised, and agreed next step.
Use "not discussed" for anything missing. Do not infer.

5. Draft a follow-up after a call

Write a short follow-up email to [SELLER NAME] after today's call.
They said: [KEY POINTS]. Reference one specific thing they told me.
No pressure language, no "just checking in", under 100 words.

If you send it as a text instead, the seller needs to have opted in to marketing texts first; more on that below.

6. Prepare for objections

A seller said: "[OBJECTION]". Give me three ways to respond that are
honest and don't pressure them. For each, say what I'm trying to
learn by responding that way.

7. Write the buyer blast for a contract

Write a deal announcement for my cash buyers list. Property facts:
[ADDRESS AREA, BEDS/BATHS, SQFT, CONDITION, REPAIRS, ARV, PRICE,
ACCESS DATE]. Lead with price and ARV. Two versions: an email under
150 words and a text under 300 characters.

8. Summarize a contract or inspection report

Here is [DOCUMENT]. List every clause that covers [INSPECTION PERIOD /
EARNEST MONEY / ASSIGNMENT / CLOSING DATE]. Quote each one exactly and
give the section number. Then flag anything that looks unusual.

Asking for exact quotes makes it easy to check the summary against the document. Any legal question still goes to your attorney.

9. Research a market with sources

Using deep research, summarize [CITY/COUNTY]: population trend,
largest employers, recent rent trend, and any rental registration or
short-term rental rules. Cite a source for every figure and link it.
Mark anything you couldn't source.

Then open the links. Deep research is good at finding sources and still capable of misreading them.

10. Write an SOP for a VA

Here is how I do [TASK] today, in my own words: [PASTE OR DICTATE].
Turn it into a numbered SOP a new VA could follow on day one. Add a
"stop and ask" line wherever a judgment call is needed.

What ChatGPT for real estate gets wrong

Comps are the big one. Ask for "recent sales near 123 Main St" with no data attached and you can get addresses, prices and dates that look real and aren't. Bring your comps from PropStream, the MLS or your title company, paste them in, and let the model format and compare them.

Local rules come second. Rental registration, short-term rental bans and disclosure requirements change by city, and a model's answer may be years out of date. Use deep research for the source, then read the source.

Listing and marketing copy is the third. The Fair Housing Act bars housing ads that indicate a preference based on a protected class (42 U.S.C. 3604(c)), and that applies to copy a model wrote the same as copy you wrote. Read every draft for wording about who should live there instead of what the house is.

And texting: prompt 5 can write the message, but a marketing text to a seller needs their prior express written consent. A2P 10DLC registration is a carrier requirement, separate from consent. The AI SMS compliance guide covers both. This is not legal advice.

When ChatGPT stops being enough

Once you're pasting the same prompt thirty times a day, the job has outgrown a chat window. Prompt 4, for example, is exactly what a voice or chat agent does on its own at the end of every call, writing the result straight into your CRM. That's the point where an agent or an automation does the work. What an AI real estate agent actually does explains that step, and how to use AI for real estate investing puts it in order.

If you're connecting tools rather than building agents, the small business automation guide using Make and the rundown of top business automation solutions are the practical starting points. For underwriting products built specifically for investors, see AI property analysis and underwriting tools. And for the version of this advice that applies outside real estate, see AI for business: a guide by function.

If you want help turning these prompts into agents that run on their own, that's what our AI agency for real estate investors builds.