How real estate agents can use ChatGPT and Claude day to day

ChatGPT and Claude are most useful to real estate agents as fast first-draft writers and thinking partners: listing descriptions, market updates, follow-up messages, lead summaries and appointment prep. They save the most time when you give them specific facts and treat what comes back as a draft you check, not a finished product.

Below are the everyday jobs worth handing over, prompts you can copy, the limits that matter, and what changes when an assistant can reach your CRM instead of just writing about it.

What are the best everyday uses of AI for real estate agents?

The best uses are writing tasks where you already know the facts and need words fast. The assistant supplies structure and phrasing; you supply the truth.

Listing descriptions

Give the assistant facts, not adjectives. The more specific your input, the less generic the output.

A prompt that works:

Write a listing description of about 150 words for a 3-bed, 2-bath single-story home. Facts (example): 1,850 sq ft, renovated kitchen with quartz counters (2024), new roof, fenced backyard with mature oak, two-car garage, walking distance to a park and a coffee shop. Tone: warm and plain, no clichés like “nestled” or “boasts.” Describe the home and its features only. Do not describe who the ideal buyer is.

That last line matters, and the fair housing section below explains why. Then ask for two alternatives: a shorter one for a portal and a two-line version for social.

Market update drafts

AI is good at turning numbers into a readable note. It is not a source of numbers. Paste in figures from your MLS or a local report you trust, and tell it to use only those.

Here are this month’s figures for our area from my MLS report: [paste]. Write a short, friendly market update email for past clients. Use only these numbers, do not add any others. Explain what they mean for someone thinking of selling in the next year, in plain language. End with an offer to run a free home value estimate.

Then check every figure in the draft against your report. If a number appears that you did not give it, delete it.

Follow-up message drafts

Assistants are useful for getting unstuck on follow-up, especially the awkward ones: a lead who went quiet, a seller who interviewed you and chose someone else, a buyer who paused their search.

A buyer lead asked about a 3-bed listing on Oak Street two weeks ago and hasn’t replied since. Write three short text messages I could send, each under 300 characters. Each should mention the Oak Street home or similar homes nearby, offer one useful thing, and make replying easy. Sound like a friendly person on a real estate team, not a marketer. No exclamation marks.

Pick one, edit it into your own voice, and send it. The draft is a starting point, not a script.

Summarizing a lead’s history

When a lead calls back after months, you want the story in thirty seconds. Paste in your notes and message history (with care; see the limits below) and ask:

Summarize this lead’s history in five bullet points: what they’re looking for, budget or timing if stated, what we’ve sent them, their last reply, and open questions. Only use what’s in the notes. If something isn’t stated, say “unknown.”

The instruction to say “unknown” helps reduce the chance of the model filling gaps with guesses.

Preparing for a listing appointment

Use the assistant as a sparring partner before you walk in.

I have a listing appointment tomorrow. The sellers have lived in the home 20 years and want to close before summer. Here are the property details and my notes: [paste]. List the ten questions or objections they’re most likely to raise, and draft a short, honest answer to each. Flag anything I should verify before the meeting.

It will not price the home for you, and it should not. What it does well is help you rehearse, organize your talking points and spot the questions you had not thought of.

What are the limits of using ChatGPT and Claude as an agent?

The limits are about privacy, accuracy and fair housing. None of them is a reason to avoid these tools, but all of them are reasons to stay in charge of what goes out.

Be careful with private client data

Share only what the task needs. A lead summary rarely needs a phone number, and nothing needs a Social Security number, bank statement or pre-approval letter pasted into a chat.

Check the data settings of the tool and plan you use. For example, Anthropic says it will train new models using data from Claude Free, Pro and Max accounts only when the user’s setting for this is on, and that this does not apply to services under its commercial terms. Other tools have their own settings and terms; read them before you paste.

Check every fact

AI tools can state things that are wrong with total confidence: a school that is not in the district, a feature the home does not have, a statistic nobody published. You are responsible for what you send. NAR’s Code of Ethics asks REALTORS to “present a true picture in their advertising, marketing, and other representations”, and that applies to AI-drafted copy as much as anything you write yourself.

Keep the language fair-housing safe

The federal Fair Housing Act makes it unlawful to publish any notice, statement or advertisement about the sale or rental of a dwelling that indicates “any preference, limitation, or discrimination based on race, color, religion, sex, handicap, familial status, or national origin”. NAR’s Code of Ethics Article 10 goes further for its members, barring denial of equal professional services on the basis of those characteristics and also sexual orientation and gender identity. State and local laws may add more.

AI models can slip into describing an imagined buyer rather than the property. A few habits help:

• Describe the property, not the people. Square footage, layout, features, distance to amenities.
• Tell the model the rule. Add “Describe the home only. Do not describe or imply who should live here” to your prompts.
• Read every draft before it goes out, especially anything about neighborhoods, schools or who a home “suits.”

What changes when your AI assistant is connected to your CRM?

Everything above is drafting. You copy, paste, edit and send. The assistant writes about your leads, but it cannot see them or act on them.

That changes when the assistant can connect to your tools. Claude and ChatGPT both support the Model Context Protocol, which describes itself as “an open-source standard for connecting AI applications to external systems”, so assistants can reach data and take actions on a user’s behalf.

This is what the Forward Flow connector for Claude and ChatGPT is for. With it, you can run your follow-up, your calendar and your CRM from the same chat where you draft your listing copy. Forward Flow itself writes every lead its own campaign of calls, texts, emails and voicemail, runs it at the hours that lead answers, and hands a lead to you live when they are ready to talk. The assistant stops being a writing tool that sits next to your leads and becomes a way to work them directly.

If you would rather not open a chat at all, the same work runs by text through Flo, Forward Flow’s assistant.

Try it from the lead’s side

The quickest way to judge AI follow-up is to receive it. Play a buyer or seller and get the real texts, voicemail and an AI phone call, or book a 20-minute demo on your own leads.

This is general information, not legal advice. Talk to a lawyer about your own situation.