Most guides to AI for real estate agents open by telling you the industry is being transformed. Then they list forty tools and leave you exactly where you started.
This one assumes you have never used any of it, are not especially interested in the technology, and mainly want to know whether it will save you time on a Tuesday afternoon. No jargon that is not explained, no tool list, and a prompt formula you can use on your next listing.
What these tools actually are
ChatGPT, Gemini, Claude and Copilot are all the same kind of thing underneath: a program trained on an enormous amount of written text, which predicts what words should come next.
That sentence is worth reading twice, because everything useful and everything dangerous about these tools follows from it.
Because it predicts language, it is extraordinarily good at anything that is fundamentally a writing problem. Rewriting, summarising, changing tone, drafting the email you have written two hundred times. It never has a blank page, never gets tired at nine in the evening, and never resents doing the boring version.
And because it predicts language, it has no idea whether what it says is true. It is not looking anything up. Ask it about a specific property and it will produce the kind of sentence that usually appears in listings for properties like that, which is not the same as a fact. This is the single most important thing to understand before you start, and it explains almost every horror story you have heard.
A useful way to hold it in your head: treat it as a fast, tireless, slightly overconfident assistant on their first week. Brilliant at drafting. Never to be trusted with a number.
Which one should you use?
Short answer: whichever one you will actually open. The differences between the major assistants matter far less than the difference between using one every day and using none.
If you want the safe default, use ChatGPT. It is what most agents use, which means every prompt guide, every YouTube walkthrough and every colleague who can help you is talking about the same interface.
Rough pricing for the individual plans, compiled from a June 2026 comparison. These change often, so check the vendor page before you subscribe:
- ChatGPT Plus around $20 a month, with a cheaper Go tier around $8.
- Google AI Pro around $20, with a lighter AI Plus tier around $5.
- Microsoft Copilot Pro around $20, with no cheaper paid option.
- Free tiers exist for all of them, capped by message or compute allowances that refresh every few hours.
Start free. The limits are generous enough for a few listing descriptions and a batch of emails, and you will learn more in two weeks of free use than in two hours of reading comparison tables. Upgrade when you hit the cap often enough to be annoyed, which is the only honest signal that the subscription is worth it.
The prompt formula
Almost everyone who says AI is useless typed one vague line and judged the result. Almost everyone who finds it useful stumbled into the same four-part structure.
Role. Context. Task. Constraints.
Here is the version most people type first:
Write a listing description for a 3 bedroom house in Boise.
You will get three paragraphs about charm and character that could describe any house in North America. Now the same request with the four parts in place:
You are an experienced listing agent writing MLS remarks.
The property is a 1978 single-storey ranch in the Boise North End, 1,480 sq ft, 3 bed, 2 bath, on a 7,000 sq ft corner lot. Original oak floors under carpet in two rooms. New roof in 2024. Kitchen is dated. Detached single garage. Two blocks from Camel’s Back Park. Likely buyer is a first-time buyer who is comfortable doing cosmetic work.
Write MLS remarks of 140 words or fewer.
Describe the property, never the buyer. No fair housing language about who it suits. Do not invent features I have not listed. Lead with the lot and the location. Mention the dated kitchen honestly rather than hiding it.
The second prompt takes ninety seconds to write and produces something you can edit rather than rewrite. Notice how much of it is constraints. That is not padding: the instruction not to invent features is what stops the tool describing a workshop that does not exist, and the fair housing line is doing real compliance work.
Save that prompt. Next listing, change the middle section and reuse everything else. Within a month you will have five or six of these, and that private collection is worth more than any list of tools.
Your first week
Five days, one task each, about twenty minutes a day. The goal is not to become proficient. It is to find out whether this saves you time on the specific work you actually do.
- Monday. Take a listing you have already published and ask for three alternative versions of the remarks using the formula above. Compare them to what you wrote. You are calibrating, not publishing.
- Tuesday. Paste in an email you find hard to write, the price reduction conversation is the classic, and ask for a version that is direct but not cold.
- Wednesday. Feed it a long document you have been avoiding, an inspection report or an HOA packet, and ask for the five things a buyer would want flagged. Then check its summary against the document.
- Thursday. Ask it to play a seller who believes their home is worth 15% more than the comps support, and to argue back the way that seller would. Practise your response.
- Friday. Generate a month of social captions in one sitting. Keep the six you would actually post.
At the end of the week you will know which of those five felt like a shortcut and which felt like homework. Keep the shortcuts. Drop the rest without guilt.
Five mistakes that cost people the first month
Publishing the first draft. The output is a starting point. Agents who paste it straight into the MLS are the reason the rest of the industry is suspicious of this.
Trusting numbers. Anything arithmetic, comps, price per square foot, net sheets, belongs in your MLS or CMA software. There is a longer explanation of why in the piece on what AI does well and where it fails, but the short version is that these tools do not calculate. They imitate the look of a calculation.
Ignoring fair housing. Phrases like “perfect for young professionals” or “ideal for families” appear in AI drafts constantly, because they appear constantly in the listing copy it learned from. They describe the buyer instead of the property, and that is a compliance problem regardless of who typed it.
Pasting client data into a public chatbot. Names, addresses, financial details. Treat the chat box as a semi-public place and describe the property without identifying the people.
Buying seven tools in week one. Every specialist product wants to be your first AI purchase. Learn what the general assistant can do, find the specific thing it handles badly, then buy the tool that fixes that. In the other order you pay for features you never open.
Questions people ask before starting
Do I need to be technical?
No. If you can write a text message you can use these tools. The skill that separates people who get good results from people who do not is not technical: it is being specific about what you want, which is the same skill that makes a good listing brief.
Can it see my MLS?
No. A general assistant has no connection to your MLS, your CRM or your local market data. Anything it says about a specific property or a specific market came from you or was invented. Some specialist real estate products do connect to those systems, and that is exactly what you are paying them for.
Will my broker have a problem with this?
Ask before you build a workflow on it. Many brokerages now have written AI policies covering client data, disclosure and what may be published. Finding out afterwards is the expensive order.
How long before it saves me real time?
For one narrow task, about two weeks. That is roughly how long it takes to build a prompt you trust and stop rewriting the output from scratch. Agents who report saving several hours a week generally got there by doing one task well and then adding a second, not by adopting everything at once.
Sources
- Pricing figures compiled from a June 2026 comparison of ChatGPT, Gemini and Copilot consumer plans. Vendor pricing changes frequently; verify before subscribing.
- HousingWire, practical uses of ChatGPT for real estate agents, including prompt structure guidance.